{"auto_groups":["default","gemini-vertex","Claude Max","Claude-G1"],"data":[{"model_name":"gpt-4-vision-preview","description":"OpenAI GPT-4 视觉预览模型，在 GPT-4 文本能力基础上增加图像理解，可用于图像问答、文档分析和视觉内容识别。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"],"pricing_version":"c576c64d115cf2ed5711aff7006a5d7ae1b7f12e6a5bbd858b36d51fa0601112"},{"model_name":"gemini-3.1-pro-preview-customtools","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该版本针对自定义工具调用进行了路由配置，适合需要外部函数或业务工具协作的智能体。","tags":"文本","vendor_id":6,"created_time":1771744664,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G2","gemini-vertex","C2","G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-5","description":"Claude Opus 5 是由 Anthropic 于 2026 年 7 月 24 日发布的旗舰级人工智能模型。该模型专注于复杂任务的代理式（agentic）编码与深度推理，具备出色的代码库理解能力，能够高效处理长期运行的多步骤任务与企业级工作流。","tags":"文本","vendor_id":2,"created_time":1785053602,"launch_time":1784880777,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["default","A2","B1","C2","Claude Max","Claude-Aws","G1","G2","A1"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-20250514","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1762411850,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default","A1","A2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-2026-04-02","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该名称固定到 2026 年 4 月 2 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1775140270,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-max-search","description":"Qwen3.7-Max 是阿里巴巴于 2026 年 5 月推出的旗舰级语言模型，专为智能体（Agent）场景设计。该模型在代码编写、办公自动化及长周期任务自主执行方面表现突出，支持高达 100 万 token 的上下文窗口。目前该模型主要通过阿里云 Model Studio 提供 API 接口服务，被广泛应用于复杂的工程开发与生产力工具构建。","tags":"文本","vendor_id":4,"created_time":1784126243,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.1,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"deepseek-chat","description":"DeepSeek 通用对话接口的稳定模型别名，面向中文与英文问答、代码生成、内容处理和工具调用，具体底层版本由服务端路由。","tags":"文本","vendor_id":3,"created_time":1785210665,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["B1","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-preview-thinking-*","description":"Google Gemini 3 Pro 旗舰多模态推理模型，擅长复杂编码、科学推理、长上下文和智能体任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为预览版本，能力和行为可能随服务商更新而调整。名称保留渠道提供的字面量 * 后缀，用于其自身的路由约定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["gemini-vertex","C2","G1","G2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5.1-codex-high","description":"OpenAI GPT-5.1-Codex高性能升级版，提供最先进的代码理解和生成能力。在GPT-5.1基础上优化，适合企业级应用、复杂系统架构和高标准代码质量要求。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["B1","default","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwq-plus","description":"阿里通义推理优化大模型，采用先进深度学习架构，支持128K上下文。在NLP任务中表现卓越，特别擅长处理复杂语言结构和生成高质量文本。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.8,"model_price":0,"owner_by":"","completion_ratio":2.5,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"net-o1-mini-2024-09-12","description":"OpenAI o1-mini模型2024年9月12日初始发布版本，提供稳定的STEM领域推理能力。适合需要版本一致性的编程任务、数学问题和科学推理应用。当前名称为服务侧路由别名，底层模型为 o1-mini-2024-09-12。该名称固定到 2024 年 9 月 12 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-chat-latest","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6","description":"Claude Opus 4.6 是 Anthropic 于 2026 年 2 月 6 日发布的最新旗舰 AI 模型‌，支持 100 万 token 上下文窗口和智能体协作功能，在编程、金融分析及多任务处理能力上实现突破性升级。‌‌\n","tags":"文本","vendor_id":2,"created_time":1770556677,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["C2","Claude Max","A1","Claude-Aws","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"grok-4.6","description":"Grok 4.6 是 xAI 于 2026 年 8 月发布的前沿 AI 模型，专为编码、智能体任务及复杂知识工作设计。该模型通过增强的推理能力和多步骤工作流处理，显著提升了在大型代码库维护及长周期工程项目中的表现。目前，Grok 4.6 已通过 xAI API、Cursor 及 GitHub Copilot 等平台提供服务，是当前 xAI 旗下的旗舰级推理模型。","tags":"文本","vendor_id":8,"created_time":1787191194,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.25,"enable_groups":["C2","codex渠道-gpt","default","G1","G2","A1","A2","B1"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-search-thinking","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"联网","vendor_id":4,"created_time":1775140237,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-pro-ci","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务。该名称为服务侧 CI 路由别名，具体底层版本由渠道配置决定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"grok-4","description":"xAI Grok 4 旗舰通用模型，面向复杂数学、科学、代码和智能体任务，支持长上下文、工具调用与实时信息处理。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1762411851,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-all","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-5","description":"Claude Sonnet 4.5 是 Anthropic 于 2025 年 9 月发布的高性能推理模型，主要面向复杂的长链路任务、智能体构建及专业代码编程场景。该模型支持 20 万上下文窗口及多模态视觉输入，曾以卓越的逻辑推理和工具使用能力在编程基准测试中表现突出。随着技术迭代，目前该模型已由更新的版本（如 Sonnet 4.6 及后续系列）接替。","tags":"文本","vendor_id":2,"created_time":1785053656,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["A1","A2","B1","default"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"grok-4.3","description":"Grok 4.3 是 xAI 于 2026 年 4 月发布的旗舰级推理大模型，基于 Colossus 超级计算机集群训练，具备百万级 Token（1M）超长上下文窗口与原生视频输入能力（支持 mp4/mov/webm 最长 5 分钟 1080p 视频的理解、转录与对象追踪），并新增 PDF、表格、幻灯片等结构化文档生成功能；该模型采用常驻/可配置推理模式（支持 none/low/medium/high 多级推理力度），在幻觉率控制、智能体工具调用及指令遵循方面表现领先，定价为输入 $1.25/百万 Token、输出 $2.50/百万 Token，支持文本、图像及视频多模态输入与 OpenAI 兼容 API，广泛应用于学术研究、代码分析、企业级合同审查及实时信息检索等场景，延续了\n《银河系漫游指南》式的幽默开放对话风格。","tags":"文本","vendor_id":8,"created_time":1784110170,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.16,"enable_groups":["A1","A2","B1","C2","codex渠道-gpt","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-flash-latest-thinking-*","description":"Google Gemini Flash 系列高速多模态模型，兼顾推理、长上下文、工具调用和成本效率，适合实时及高并发应用。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。名称保留渠道提供的字面量 * 后缀，用于其自身的路由约定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5-codex-high","description":"OpenAI GPT-5-Codex高性能版本，提供更强的代码理解和生成能力。适合复杂代码重构、架构设计和大规模系统开发，在代码质量和智能程度上超越标准版本。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v4-1-flash-260910","description":"DeepSeek 于 2026 年 9 月 10 日推出 DeepSeek-V4.1-Flash，是一款 552B 参数的高性价比多模态混合专家（MoE）模型。该模型原生支持视觉理解与百万级上下文，大幅压缩了 KV 缓存，专为长流程智能体及高并发场景提供极低成本的高效推理支持。","tags":"文本","vendor_id":3,"created_time":1789723933,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"(tier(\"base\", p * 1 + c * 4 + cr * 0.02)) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 9 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 12 ? 2 : 1) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 14 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 18 ? 2 : 1)"},{"model_name":"gemini-3.1-flash-image-2k","description":"Gemini 3.1 Flash Image（代号 Nano Banana 2）是 Google 于 2026 年 6 月正式发布的图像生成与编辑模型。该模型专注于高效率与低延迟，支持通过自然语言进行对话式图像编辑、多角色一致性保持及实时参考引用，主要面向高频次、大规模的图像生成与创意工作流场景。","tags":"图片","vendor_id":6,"created_time":1785055004,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"qwen-turbo-2025-04-28-search-thinking","description":"通义千问Turbo 2025年4月28日版本联网搜索思考模式。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-flash-2025-10-15-thinking","description":"轻量级多模态视觉-语言模型2025年10月15日版本思考模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-1-thinking-all","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。当前名称为按次计费别名，实际模型名称为 claude-opus-4-1-thinking-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"grok-4-1-fast-reasoning","description":"xAI Grok 4.1 高能力通用模型，强化推理、代码、工具调用和智能体任务，在复杂工作流中兼顾准确性与执行效率。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本针对响应速度和吞吐量进行了优化。","tags":"文本","vendor_id":8,"created_time":1785210665,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":2.5,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-2024-08-06","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。该名称固定到 2024 年 8 月 6 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-chat","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-image-2","description":"GPT Image 2 是 OpenAI 于 2026 年 4 月推出的新一代图像生成模型。该模型采用全新的自回归架构，在图像生成速度、真实感及多语言文本渲染精度上均有显著提升，文本渲染准确率高达 99%。它主要面向专业设计与商业应用场景，支持高保真图像生成与精准的局部编辑，被广泛集成于多种创作工具与 API 平台中。","tags":"图片","vendor_id":1,"created_time":1776761956,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["A1","A2","B1","default"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-fast","description":"Google Gemini 3.1 系列快速路由模型，面向低延迟对话、代码辅助和高吞吐智能体任务，具体底层版本由渠道配置决定。该版本针对响应速度和吞吐量进行了优化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"grok-3-reasoner","description":"xAI Grok 3 通用大语言模型，擅长推理、代码、实时信息分析和工具使用，适合对话、研究与复杂问题求解。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"kimi-k3","description":"月之暗面 Kimi K3 高能力大语言模型，面向复杂推理、代码、长上下文和智能体任务，适合多步骤生产工作流。","tags":"文本","vendor_id":9,"created_time":1785210665,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A2","B1","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-latest","description":"通义千问Turbo最新版本，优化的快速响应模型。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-flash-search-thinking","description":"轻量级多模态视觉-语言模型联网搜索思考模式，快速响应结合联网和视觉理解。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-flash-preview","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。","tags":"文本","vendor_id":6,"created_time":1766546328,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["A1","A2","default","G1","G2","B1","C2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-preview-thinking","description":"通义千问旗舰模型预览版思考模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.4-pro-2026-03-05","description":"OpenAI GPT-5.4 系列高能力专业模型，面向高难度推理、软件工程、研究和长程智能体任务，强调准确性与任务完成质量。该名称固定到 2026 年 3 月 5 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1773730258,"quota_type":0,"model_ratio":15,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o3-pro-2025-06-10","description":"OpenAI o3-pro模型2025年6月10日官方发布版本，标志着史上最强AI模型的正式上线。提供完整的工具访问能力和最高推理性能，适合最复杂的专业应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-3-mini","description":"xAI Grok 3 Mini 轻量推理模型，兼顾速度、成本与逻辑分析能力，适合数学、代码和高并发问答。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1762411851,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":1.67,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-latest-thinking","description":"通义千问Turbo最新版思考模式，优化的快速响应结合推理能力。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-plus-search-thinking","description":"多模态视觉-语言模型联网搜索思考模式，结合视觉理解和联网搜索能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-thinking","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":4,"created_time":1775140257,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro-thinking","description":"Google Gemini 2.5 Pro思考版，启用可见的推理过程展示。通过显示模型的思考步骤，帮助用户理解复杂决策逻辑。具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。配备\"Deep Think\"增强推理模式，专门针对复杂数学和编码问题进行优化。","tags":"文本","vendor_id":6,"created_time":1766666140,"quota_type":0,"model_ratio":0.628,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","C2","default","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"suno_act_timing","description":"Suno AI时间戳歌词功能，生成带有精确时间戳的歌词文本。每句歌词标注对应的播放时间点，适合制作字幕文件、卡拉OK显示、歌词同步展示等应用场景。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.4-nano","description":"OpenAI GPT-5.4 系列最轻量模型，面向分类、抽取、简单生成和大规模低延迟处理，强调速度与成本效率。","tags":"文本","vendor_id":1,"created_time":1773982932,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":6.25,"cache_ratio":0.1,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-4-0709","description":"xAI Grok 4 旗舰通用模型，面向复杂数学、科学、代码和智能体任务，支持长上下文、工具调用与实时信息处理。该名称为指定发布批次的固定版本，便于保持行为一致性。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-4.20-0309-reasoning","description":"grok-4.20-0309-reasoning 是 xAI 于 2026 年 3 月 9 日发布的 Grok 4.20 系列的推理（Reasoning）变体，它在生成最终回复前会执行扩展的链式思维（Chain-of-Thought）内部推导过程以提升准确性，专为复杂逻辑、数学解题、代码调试、科学研究及多步调查等高准确度要求的任务设计，具备 100 万 token 上下文窗口、多模态（文本+图像）输入、原生函数调用与结构化输出能力，在牺牲一定首 token 延迟（平均约 10.85 秒）并消耗额外推理 token 成本的前提下，可将幻觉率较前一版本降低约 65%，API 定价与同系列非推理版一致（输入 $1.25/百万 token、缓存 $0.20/百万 token、输出 $2.50/百万 token），适合在准确性优先于速度的严肃场景中使用。","tags":"文本","vendor_id":8,"created_time":1784110045,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.16,"create_cache_ratio":0.16,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-3.5-turbo","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G2","A1","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.2-chat-2025-12-11","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。该名称固定到 2025 年 12 月 11 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G1","G2","A1","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-sonnet-4-thinking-all","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-thinking-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"grok-4.20-beta-0309-reasoning","description":"xAI Grok 4.20 Beta 0309 推理版本，面向复杂分析、代码、研究和长程智能体任务，支持约 200 万 token 长上下文并启用增强推理；该名称固定到 2026 年 3 月 9 日版本。","tags":"文本","vendor_id":8,"created_time":1786443762,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.16,"create_cache_ratio":0.16,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-flash-search","description":"通义千问Flash联网搜索版本，快速响应结合实时联网能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-2025-04-28-thinking","description":"通义千问Turbo 2025年4月28日版本思考模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-haiku-4-5-20251001","description":"Anthropic Claude Haiku 4.5 高速高性价比模型，强化编码、智能体和工具调用能力，适合实时交互与大规模生产工作负载。该名称固定到 2025 年 10 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1762411851,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["G1","G2","A1","A2","B1","Claude Max","Claude-Aws","default","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-deepsearch","description":"Google Gemini 3 Pro深度搜索版，集成增强的搜索和研究能力。利用最先进的推理能力处理多信息源（文本、音频、图像、视频、PDF、代码库），进行深度信息检索、分析和综合。支持1百万token上下文窗口，适合需要全面调研、多源信息整合和深度分析的复杂研究任务。","tags":"研究","vendor_id":6,"created_time":1766666183,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","A1","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"qwen-turbo-latest-search-thinking","description":"通义千问Turbo最新版联网搜索思考模式，快速响应结合联网搜索和推理。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-6-c","description":"Anthropic Claude Sonnet 4.6 高性能均衡模型，面向复杂编码、长程智能体、工具使用和专业知识工作，在速度与能力之间取得平衡。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-6。","tags":"文本","vendor_id":2,"created_time":1771926464,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"cs-claude-opus-4-all","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-turbo","description":"OpenAI GPT-4 Turbo 模型，在保持较强推理与代码能力的同时提供更长上下文、更高速度和更低成本，支持文本与视觉任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-flash-latest","description":"Google Gemini Flash 系列高速多模态模型，兼顾推理、长上下文、工具调用和成本效率，适合实时及高并发应用。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"qwen3-max-preview","description":"通义千问旗舰模型预览版本，提供最新功能体验。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-plus-2025-12-19-thinking","description":"多模态视觉-语言模型2025年12月19日版本思考模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-audio-3.0-asr-flash-filetrans","description":"全面支持汉语传统七大方言体系（官话/吴/湘/赣/客/闽/粤），并适配 20+ 地区口音官话。针对中文古诗词的韵律、节奏与文言表达特点进行专项优化，提升对古诗词内容的识别准确率，适用于文化传承、教育讲解、有声读物等场景。优化标点预测与文本归一化能力，使输出文本更符合书面表达习惯，数字、日期、金额等信息自动转换为标准格式，增强内容的可读性与专业性。同时语种扩展至英语、日语、韩语、越南语、泰语、印尼语、马来语、菲律宾语、印地语、阿拉伯语、法语、德语、西班牙语、葡萄牙语、俄语、意大利语、荷兰语、瑞典语、丹麦语、芬兰语、挪威语、希腊语、波兰语、捷克语、匈牙利语、罗马尼亚、保加利亚语、克罗地亚语、斯洛伐克语等，共计30个语种。支持context上下文能力，可转写5分钟以内的音频。","tags":"语音","vendor_id":4,"created_time":1789459182,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI语音转文本"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-medium","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该条目将推理强度设为中档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6","description":"GPT-5.6 系列通用模型，面向复杂编码、研究分析、工具调用和多步骤智能体任务，兼顾高质量推理与稳定的生产级输出。","tags":"文本","vendor_id":1,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","A2","B1","C2","codex渠道-gpt","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5 + cc * 6.25) : tier(\"long_context\", p * 10 + c * 45 + cr * 1 + cc * 12.5)"},{"model_name":"qwen3-max-2025-09-23","description":"通义千问旗舰模型2025年9月23日版本，支持256K超大上下文。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-gpt-4o","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。当前名称为按次计费别名，实际模型名称为 gpt-4o。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"suno_music","description":"Suno AI核心音乐生成功能，一款生成式AI音乐创作程序，能产生包含人声与乐器的逼真歌曲。用户只需输入歌词、歌名或选择音乐风格，即可在一分钟内生成两首时长2-4分钟的完整歌曲。V4版本提供更清晰细腻的音频效果，高音穿透力和低音沉稳感显著提升。支持多语言、多风格，适合个人创作、音乐教育、商业制作、游戏开发等场景。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.35,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-lite-preview-09-2025","description":"Google Gemini 2.5 Flash-Lite预览版（2025年9月版本），最具成本效益和最快的2.5模型。相比之前版本具有更低的延迟，特别适用于翻译、分类等需要大规模处理的场景。保持了Gemini 2.5的核心能力，提供快速响应和高并发处理能力。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"o3-mini-low","description":"OpenAI o3-mini低推理强度版本，优先考虑响应速度。当延迟是关键考虑因素时选择此版本，提供快速推理结果，适合需要实时响应的应用场景。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G2","C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-7-thinking","description":"Anthropic Claude Opus 4.7 旗舰模型，面向高难度软件工程、复杂推理和长周期自主智能体任务，强调深度分析与可靠执行。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1776388557,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.7-flash","description":"Gemini 3.7 Flash 是谷歌于 2026 年 8 月发布的 Flash 系列最新模型，被定位为面向代码开发、智能体构建及复杂业务流程的“工作主力”模型。该模型在软件工程、Web 开发及金融、法律等知识密集型领域表现出色，相较于前代产品在逻辑推理、指令遵循及多步任务规划能力上均有显著提升，旨在为开发者提供高效且具成本效益的生产级 AI 支持。","tags":"文本","vendor_id":6,"created_time":1786853859,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":2.5,"cache_ratio":0.1,"enable_groups":["G1","A2","B1","C2","G2","gemini-vertex","A1","default"],"supported_endpoint_types":["gemini","openai","Gemini"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6-thinking-c","description":"Claude Opus 4.6 思考版 是 Anthropic 于 2026 年 2 月 6 日发布的最新旗舰 AI 模型‌，支持 100 万 token 上下文窗口和智能体协作功能，在编程、金融分析及多任务处理能力上实现突破性升级。‌‌。当前名称为按次计费别名，实际模型名称为 claude-opus-4-6-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1771926475,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-4","description":"OpenAI GPT-4 系列高能力通用模型，擅长复杂指令遵循、逻辑推理、代码生成和专业文本处理，是早期 GPT-4 代旗舰模型。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":15,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.5,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-pro","description":"OpenAI GPT-5 系列专业高能力版本，面向复杂推理、科研分析、软件工程和高价值知识工作，强调答案质量与可靠性。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"happyhorse-1.0-r2v-1080P","description":"参考生视频","tags":"视频","vendor_id":4,"created_time":1780043294,"quota_type":1,"model_ratio":0,"model_price":1.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"claude-sonnet-4-20250514-thinking","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":2,"created_time":1766657046,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"cs-grok-4","description":"xAI Grok 4 旗舰通用模型，面向复杂数学、科学、代码和智能体任务，支持长上下文、工具调用与实时信息处理。当前名称为按次计费别名，实际模型名称为 grok-4。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.02,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-flash-lite-image-1k","description":"Gemini 3.1 Flash-Lite Image 是 Google Gemini 3.1 系列中专注于视觉生成的轻量化多模态模型。该模型以极高的推理速度和成本效益为核心设计，主要面向高频、大规模的图像生成与编辑任务。它支持文本与图像间的交互生成，具备出色的提示词遵循能力，常被用于快速创意迭代及移动端高效素材制作等场景。\n","tags":"图片","vendor_id":6,"created_time":1785830269,"quota_type":1,"model_ratio":0,"model_price":0.12,"owner_by":"","completion_ratio":0,"enable_groups":["default","Gemini-image-G1","B1"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI生图","OpenAI改图"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v3.2-exp","description":"DeepSeek-V3.2-Exp 实验模型，引入稀疏注意力等效率优化，面向长上下文、代码、推理和智能体任务验证新架构能力。","tags":"文本","vendor_id":3,"created_time":1766650015,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["A1","A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro","description":"Google于2025年6月正式发布的旗舰级AI模型，具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。支持原生音频输出、高级安全防护和Project Mariner的计算机使用功能。配备\"Deep Think\"实验性增强推理模式，专门针对复杂数学和编码问题进行优化。最大输入1,048,576 tokens，输出65,535 tokens，支持多模态输入和丰富的工具集成能力。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","G1","G2","B1","C2","gemini-vertex","A1","A2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-flash-thinking","description":"通义千问Flash思考模式，快速响应的同时支持复杂推理分析。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"omni_flash-10s","description":"Omni Flash（10s） 是Gemini一款主要用于视频生成的 AI 模型，常见于各类 AI 聚合平台和 API 服务中。该模型支持文生视频及图生视频等功能，常被用于产品展示、短剧制作及创意内容创作，能够根据用户提供的提示词或参考图像，快速生成特定时长的视频片段。","tags":"视频","vendor_id":6,"created_time":1779851802,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"grok-3-fast-beta","description":"xAI Grok 3 Fast Beta 高速通用模型，提供 131K 上下文窗口，面向实时对话、编码、工具调用和智能体任务，强调低延迟与高吞吐。","tags":"文本","vendor_id":8,"created_time":1786443762,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-4.20-beta-0309-non-reasoning","description":"xAI Grok 4.20 Beta 0309 非推理版本，面向低延迟对话、代码、工具调用和高吞吐智能体任务，支持约 200 万 token 长上下文；该名称固定到 2026 年 3 月 9 日版本。","tags":"文本","vendor_id":8,"created_time":1786443762,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.16,"create_cache_ratio":0.16,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-codex-medium","description":"OpenAI GPT-5.1-Codex中等性能版本，在5.1系列中提供最佳的性能价格比。适合标准企业开发、团队协作编码和中等复杂度的项目。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-image-4k","description":"Google Gemini 3 Pro Image预览版支持打印级质量输出。具备思考推理能力和搜索Grounding功能，可生成动态内容。支持12+种语言的多语言文字渲染，高级图像混合可融合至14张图像，专注于复杂、多轮、高保真的专业级生成任务。","tags":"图片","vendor_id":6,"created_time":1766546719,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-low","description":"Google Gemini 3 Pro低推理强度版，优化响应速度和成本效率。通过thinking_level参数控制推理深度，提供快速而准确的响应。保留Gemini 3 Pro的核心能力，但减少推理时间，适合需要快速响应的常规任务、实时对话和高频查询场景。","tags":"文本","vendor_id":6,"created_time":1766666222,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"glm-5","description":"GLM-5是由中国领先的人工智能公司智谱人工智能（Z.ai）开发的第五代大型语言模型。该模型代表了人工智能能力的飞跃，采用混合专家（MoE）架构，拥有约7450亿个总参数，每次推理活跃参数达440亿。GLM-5从底层架构开始构建，旨在实现智能体智能、高级多步骤推理，并在编码、创意写作和复杂问题解决等领域展现前沿性能。","tags":"文本","vendor_id":7,"created_time":1770863863,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4.5,"cache_ratio":0.25,"enable_groups":["C2","default","G1","G2","A1","A2","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-chat-2025-08-07","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该名称固定到 2025 年 8 月 7 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-terra","description":"GPT-5.6 Terra 随 GPT-5.6 系列于 2026 年 6 月 26 日 一同发布，定位为能力、速度和成本之间的均衡型模型。它适合日常办公、内容生成、数据分析、代码辅助、知识问答和多步骤任务处理等通用场景。相比旗舰模型 Sol，Terra 更注重成本效率；相比轻量模型 Luna，它又保留了更强的综合能力，因此适合作为多数业务场景中的默认选择。","tags":"文本","vendor_id":1,"created_time":1783660381,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","codex渠道-gpt","G1","A1","C2","default","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 2 + c * 12 + cr * 0.2 + cc * 2.5) : tier(\"long_context\", p * 4 + c * 18 + cr * 0.4 + cc * 5)"},{"model_name":"grok-4.2","description":"xAI Grok 4.2 旗舰通用模型，面向复杂推理、编码、实时搜索和长程智能体任务，强化可靠性与工具协作能力。","tags":"文本","vendor_id":8,"created_time":1773030773,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"cs-grok-4-1-fast","description":"xAI Grok 4.1 高能力通用模型，强化推理、代码、工具调用和智能体任务，在复杂工作流中兼顾准确性与执行效率。当前名称为按次计费别名，实际模型名称为 grok-4-1-fast。该版本针对响应速度和吞吐量进行了优化。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.016,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-turbo-preview","description":"OpenAI GPT-4 Turbo 模型，在保持较强推理与代码能力的同时提供更长上下文、更高速度和更低成本，支持文本与视觉任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o3-all","description":"OpenAI o3全能版本，整合最强推理模型的所有能力和工具集成。提供完整的网络搜索、Python数据分析、图像分析等功能，适合需要全面o3能力的复杂应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"net-o1-preview-all","description":"OpenAI o1-preview全能版本，提供广泛通用知识的完整推理能力。适合需要跨领域知识和博士级推理的综合性复杂问题解决。当前名称为服务侧路由别名，底层模型为 o1-preview-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"o3-mini-2025-01-31-high","description":"OpenAI o3-mini 轻量推理模型，擅长数学、科学、代码和多步骤问题求解，并可通过推理强度配置平衡速度与答案质量。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。该名称固定到 2025 年 1 月 31 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-2024-11-20","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。该名称固定到 2024 年 11 月 20 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["G1","G2","A1","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.2-chat-latest","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["gemini","openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.5","description":"OpenAI GPT-5.5 前沿通用模型，面向复杂编码、研究分析和真实业务任务，强化长链路推理、工具协作与高质量专业输出。","tags":"文本","vendor_id":1,"created_time":1776993761,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","codex渠道-gpt","G1","G2","A1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5) : tier(\"long_context\", p * 10 + c * 45 + cr * 1)"},{"model_name":"claude-opus-4-thinking-all-c","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-thinking-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1770778746,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"cs-qwq-32b","description":"阿里通义 QwQ-32B 开源推理模型，专注数学、代码和复杂逻辑问题，通过强化学习提升逐步推理与问题求解能力。当前名称为按次计费别名，实际模型名称为 qwq-32b。","tags":"文本","vendor_id":4,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-flash-image-preview-4k","description":"Nano Banana 2（官方标识为 gemini-3.1-flash-image-preview）是 Google DeepMind 发布的新一代 AI 图像生成与编辑模型。该模型基于 Gemini 3.1 Flash Image 架构打造，在保持 Flash 级别极速生成的同时，大幅缩小了与 Pro 版本在画质上的差距，实现了速度、质量与成本的高效平衡。","tags":"图片","vendor_id":6,"created_time":1772158089,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini-image-G1","B1","default"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-nano","description":"OpenAI GPT-5 系列最轻量模型，针对低延迟和大规模调用优化，适合分类、抽取、排序和简单生成任务。","icon":"OpenCode","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.025,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-search-thinking","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"文本","vendor_id":4,"created_time":1772075165,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"cs-gpt-4.1-mini","description":"OpenAI GPT-4.1 系列轻量模型，具备优秀的指令遵循、代码能力和长上下文理解，适合兼顾性能、速度与成本的应用。当前名称为按次计费别名，实际模型名称为 gpt-4.1-mini。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.005,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-preview_vertex-c","description":"Google Gemini 3 Pro 旗舰多模态推理模型，擅长复杂编码、科学推理、长上下文和智能体任务。该名称为 Vertex 路由的按次计费别名，具体底层版本由渠道配置决定。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"grok-4-fast-reasoning","description":"xAI Grok 4 旗舰通用模型，面向复杂数学、科学、代码和智能体任务，支持长上下文、工具调用与实时信息处理。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本针对响应速度和吞吐量进行了优化。","tags":"文本","vendor_id":8,"created_time":1766546328,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["G1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o4-mini","description":"OpenAI o4系列小型优化模型，专注于快速、成本高效的推理。在数学、编码和视觉任务中表现出众，在AIME 2024和2025考试中表现最佳。针对需要快速响应的应用场景优化，提供出色的性价比。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-flash","description":"Qwen3.7-flash 是阿里巴巴推出的多模态推理模型，主要面向高吞吐量的多模态任务与低成本智能体开发。该模型支持文本、图像及视频输入，具备约 100 万 token 的上下文处理能力，在视觉感知与长序列理解任务中表现出极高的性价比。作为 Qwen 3.7 系列的轻量化版本，它常被用于大规模自动化工作流与各类视觉智能代理场景。","tags":"文本","vendor_id":4,"created_time":1786442176,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 32000\n  ? tier(\"standard\", p * 0.2 + c * 0.8 + cr * 0.04)\n  : (len \u003c= 256000\n    ? tier(\"medium_context\", p * 0.6 + c * 2.4 + cr * 0.12)\n    : tier(\"long_context\", p * 1.2 + c * 4.8 + cr * 0.24))"},{"model_name":"cs-claude-opus-4","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.0-flash-lite-preview-02-05","description":"Google Gemini 2.0 Flash-Lite 轻量多模态模型，面向低延迟、低成本和高吞吐任务，支持文本、图像、音频与视频输入。","tags":"文本","vendor_id":6,"created_time":1766658129,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.4-mini","description":"OpenAI GPT-5.4 系列轻量模型，在代码、推理和工具调用能力与成本之间取得平衡，适合高并发及中等复杂度任务。","tags":"文本","vendor_id":1,"created_time":1773982919,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-2024-05-13","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。该名称固定到 2024 年 5 月 13 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-20250514-thinking","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-20250514-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-image","description":"Google Gemini 3 Pro Image标准版，基于Gemini 3 Pro构建的专业级图像生成模型。新增\"思考\"推理能力，模型先思考再生成，提升复杂提示的准确性。支持搜索Grounding集成实时数据，多语言文本渲染支持12+种语言，高级图像混合可融合多张图像。","icon":"Gemini.Color","tags":"图片","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.3-codex-spark","description":"GPT-5.3-Codex-Spark 是 OpenAI 于 2026 年 2 月推出的专用于代码生成的轻量化模型。该模型基于 Cerebras WSE-3 硬件加速，以实现超低延迟的实时编码体验，每秒可生成超过 1,000 个 token。目前主要作为研究预览版提供给 ChatGPT Pro 用户，适用于快速代码迭代、单元测试生成及代码审查等场景。","tags":"文本","vendor_id":1,"created_time":1785151238,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"create_cache_ratio":0.1,"enable_groups":["A1","A2","C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o3-mini","description":"OpenAI o3系列小型优化模型（2025年1月发布），首次向ChatGPT免费用户开放推理模型。响应速度比o1-mini提升24%，中等推理强度表现优于o1完整版。支持low、medium、high三个推理努力级别，可在准确性和延迟之间灵活权衡。在数学、编码和视觉任务中表现出众，AIME 2025通过率达99.5%。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-search","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"文本","vendor_id":4,"created_time":1772075156,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6-c","description":"Claude Opus 4.6 是 Anthropic 于 2026 年 2 月 6 日发布的最新旗舰 AI 模型‌，支持 100 万 token 上下文窗口和智能体协作功能，在编程、金融分析及多任务处理能力上实现突破性升级。‌‌。当前名称为按次计费别名，实际模型名称为 claude-opus-4-6。","tags":"文本","vendor_id":2,"created_time":1770778720,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.4","description":"OpenAI GPT-5.4 通用旗舰模型，擅长编码、复杂推理、知识工作和工具调用，适合作为专业应用与智能体工作流的核心模型。","tags":"文本","vendor_id":1,"created_time":1772778808,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 2.5 + c * 15 + cr * 0.25) : tier(\"long_context\", p * 5 + c * 22.5 + cr * 0.5)"},{"model_name":"gemini-3-flash-preview-thinking","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-1.5-flash","description":"Google Gemini 1.5 Flash 高速多模态模型，支持长上下文以及文本、图像、音频和视频理解，适合高吞吐内容处理与实时应用。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.63,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"nano-banana","description":"Google DeepMind推出的AI图像生成与编辑模型，正式名称为Gemini 2.5 Flash Image。支持文本驱动图像生成、多图融合、多轮编辑、角色一致性、风格迁移等功能。采用交错生成技术，价格仅为OpenAI同类服务的约20%。适用于营销、电商设计、内容创作、数字艺术等领域。","tags":"图片","vendor_id":6,"created_time":1763611614,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"nano-banana-2","description":"Nano Banana 2（官方标识为 gemini-3.1-flash-image-preview）是 Google DeepMind 发布的新一代 AI 图像生成与编辑模型。该模型基于 Gemini 3.1 Flash Image 架构打造，在保持 Flash 级别极速生成的同时，大幅缩小了与 Pro 版本在画质上的差距，实现了速度、质量与成本的高效平衡。","tags":"图片","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini-image-G1","B1","default"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-fast-480p-i2v","description":"支持 480p 分辨率。Seedance 2.0 Fast 高速视频生成模型，仅支持文生视频和图生视频，面向快速预览、高频迭代和低延迟创作。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.28,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-6-sol","description":"高性能性价比专业模型，主打复杂代码开发、数据分析、专业文档处理、智能体任务，API 价格相比上一代 GPT-5.6 直接减半。","tags":"文本","vendor_id":1,"created_time":1790170600,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G2","A2","C2","codex渠道-gpt","default","A1","B1","G1"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI对话（响应式）"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5)"},{"model_name":"cs-claude-sonnet-4-5-all","description":"Claude Sonnet 4.5 是 Anthropic 于 2025 年 9 月发布的高性能推理模型，主要面向复杂的长链路任务、智能体构建及专业代码编程场景。该模型支持 20 万上下文窗口及多模态视觉输入，曾以卓越的逻辑推理和工具使用能力在编程基准测试中表现突出。随着技术迭代，目前该模型已由更新的版本（如 Sonnet 4.6 及后续系列）接替。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-5-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-pro-preview_vertex-c","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务。该名称为 Vertex 路由的按次计费别名，具体底层版本由渠道配置决定。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"net-o1-preview-2024-09-12","description":"OpenAI o1-preview模型2024年9月12日初始发布版本，提供广泛通用知识的复杂问题解决能力。适合需要稳定博士级推理能力的法律分析、科研推理和金融决策等场景。当前名称为服务侧路由别名，底层模型为 o1-preview-2024-09-12。该名称固定到 2024 年 9 月 12 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"qwen-turbo","description":"通义千问高性价比快速响应模型，适合对响应速度要求较高的应用场景。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6-20260203-c","description":"Claude Opus 4.6 是 Anthropic 于 2026 年 2 月 6 日发布的最新旗舰 AI 模型‌，支持 100 万 token 上下文窗口和智能体协作功能，在编程、金融分析及多任务处理能力上实现突破性升级。‌‌。当前名称为按次计费别名，实际模型名称为 claude-opus-4-6-20260203。该名称固定到 2026 年 2 月 3 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":2,"created_time":1771926485,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"grok-code-fast-1","description":"xAI Grok Code Fast 1 高速编程模型，针对代码生成、调试、仓库检索和智能体开发流程优化，强调低延迟与工具调用效率。该版本针对响应速度和吞吐量进行了优化。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1762411851,"quota_type":0,"model_ratio":0.52,"model_price":0,"owner_by":"","completion_ratio":3.9,"cache_ratio":0.2,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.3-codex-openai-compact","description":"GPT-5.3-Codex-OpenAI-Compact 是基于 OpenAI GPT-5.3-Codex 模型架构的衍生版本，主要由第三方 API 服务商提供。该模型延续了原版在自主软件工程、终端自动化及代码审计方面的核心能力，支持 40 万 token 上下文窗口。它被广泛应用于需要高精度代码生成与复杂任务自主执行的工程开发场景，旨在通过特定配置满足开发者对高效生产力工具的需求。","tags":"文本","vendor_id":1,"created_time":1784126409,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-lite-image","description":"Gemini 3.1 Flash-Lite Image 是 Google Gemini 3.1 系列中专注于视觉生成的轻量化多模态模型。该模型以极高的推理速度和成本效益为核心设计，主要面向高频、大规模的图像生成与编辑任务。它支持文本与图像间的交互生成，具备出色的提示词遵循能力，常被用于快速创意迭代及移动端高效素材制作等场景。","tags":"图片","vendor_id":6,"created_time":1785830024,"quota_type":1,"model_ratio":0,"model_price":0.12,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["gemini","openai","Gemini"],"billing_labels":["按次计费"]},{"model_name":"claude-opus-4-8","description":"Claude Opus 4.8 是由 Anthropic 于 2026 年 5 月发布的智能模型，主要面向企业级工作流，在代码编写、多步推理及自主智能体任务中表现突出。该模型支持 100 万 token 的上下文窗口，具备较强的复杂逻辑处理与多阶段调试能力，适用于长周期、高协作的知识处理与项目编排场景。","tags":"文本","vendor_id":2,"created_time":1780017689,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","Claude Max","default","G2","A1","A2","C2","Claude-Aws","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-low","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该条目将推理强度设为低档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-2025-09-23-search","description":"通义千问旗舰模型2025年9月23日版本联网搜索。","tags":"联网","vendor_id":4,"created_time":1769691215,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v3.1","description":"DeepSeek V3.1升级版对话模型，在V3基础上进行优化改进。采用non-thinking模式提供快速响应，基于671B总参数量，每token激活37B参数。支持128K上下文窗口，在代码生成、多语言理解、指令遵循方面性能提升。适合常规对话、代码辅助和文档处理任务。","tags":"文本","vendor_id":3,"created_time":1762411850,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["B1","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"claude-fable-5","description":"Claude-Fable-5 是 Anthropic 于 2026年6月9 日发布的首个面向公众开放的 Mythos 级（超越 Opus 层级）旗舰大语言模型，也是其迄今为止公开发布能力最强的通用模型，专为长周期自主任务、复杂软件工程与高阶知识工作设计，内置安全护栏以支持更广泛的安全使用。","tags":"文本","vendor_id":2,"created_time":1781056771,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","C2","Claude Max","Claude-Aws","G2","A1","A2","default","G1"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.2-chat","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。","tags":"文本","vendor_id":1,"created_time":1767605131,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-long-latest","description":"通义千问Long最新版本，超长上下文窗口，专为长文本任务优化。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-all","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.05,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"qwen-plus-latest-thinking","description":"通义千问Plus最新版思考模式，专门用于复杂问题的逐步推理分析。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":13.333333333333,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-image-preview-4k","description":"Google Gemini 3 Pro Image预览版支持打印级质量输出。具备思考推理能力和搜索Grounding功能，可生成动态内容。支持12+种语言的多语言文字渲染，高级图像混合可融合至14张图像，专注于复杂、多轮、高保真的专业级生成任务。","icon":"Gemini.Color","tags":"图片","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini-image-G1","B1","default"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-turbo-2024-04-09","description":"OpenAI GPT-4 Turbo 模型，在保持较强推理与代码能力的同时提供更长上下文、更高速度和更低成本，支持文本与视觉任务。该名称固定到 2024 年 4 月 9 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o1-pro","description":"OpenAI o1系列最强大的推理模型，使用更多计算资源进行深层思考。上下文窗口200,000 tokens，最多生成100,000 tokens。支持视觉输入、函数调用、结构化输出。成本150美元/百万输入tokens，适合跨学科研究、代码生成优化、系统架构设计等最复杂的推理任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"qwen-plus-thinking","description":"通义千问Plus思考模式，用于复杂任务的深度推理分析。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":13.333333333333,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-search","description":"通义千问Turbo联网搜索版本，高性价比快速响应结合实时联网能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-2026-04-02-search","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。该名称固定到 2026 年 4 月 2 日版本，适合需要版本一致性的应用。","tags":"联网","vendor_id":4,"created_time":1775140167,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-low","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该条目将推理强度设为低档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-plus","description":"多模态视觉-语言模型，支持文本、图像和视频输入。支持262K超长上下文，具备2D/3D定位、空间感知、高级OCR、多语言识别等增强功能。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.8-max","description":"Qwen3.8-Max 是通义千问原生视觉语言旗舰模型，总参数约 2.4 万亿，支持图像、文本和视频输入，具备超长上下文、复杂推理、编程、办公自动化与长程智能体能力，适合端到端完成专业任务。","tags":"文本","vendor_id":4,"created_time":1786443762,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.2,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"happyhorse-1.0-r2v-720P","description":"参考生视频","tags":"视频","vendor_id":4,"created_time":1780043269,"quota_type":1,"model_ratio":0,"model_price":0.9,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["default","B1"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-3.5-turbo-0125","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。该名称固定到 2024 年 1 月 25 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":12,"cache_ratio":4,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-30b-a3b-thinking","description":"Qwen3-30B-A3B（思考型）是Qwen3系列的最新成员，采用先进的密集与混合专家（MoE）架构，创新性地实现了思考模式与非思考模式的无缝融合，支持在对话中灵活切换。尽管参数规模更为精简，其推理能力已可媲美QwQ-32B，通用能力显著超越Qwen2.5-14B，在同规模模型中达到业界SOTA水平。凭借在推理、指令遵循、智能代理和多语言支持等方面的全面突破，Qwen3-30B-A3B为高效智能应用提供了理想的解决方案。","tags":"文本","vendor_id":4,"created_time":1766666400,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"text-embedding-v4","description":"阿里通义多语言文本统一向量模型，基于Qwen3训练。支持最大128K输入长度，用户自定义64~2048维向量维度，相比V3版本提升15%~40%。","tags":"向量","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","embeddings"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-search","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.05,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"MiniMax-H3-1080p","description":"MiniMax H3 是由 MiniMax 开发的一款通用多模态视频生成模型，支持文本、图像、视频和音频的联合输入与生成。其中的 1080p 模式是该模型的一种性能优化配置，主要用于高效的创作预览、本地草稿测试或降低推理成本，在保持核心生成能力的同时提供更快的响应速度。该模型具备原生立体声输出和精准的视频编辑能力，被广泛应用于各类型的创意视频制作场景中。","tags":"视频","vendor_id":34,"created_time":1790667672,"quota_type":1,"model_ratio":0,"model_price":0.5,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["default","B1"],"supported_endpoint_types":["openai","OpenAI视频"],"billing_labels":["按次计费"]},{"model_name":"qwen-flash","description":"速度快、成本低的通义千问模型，适合简单任务。支持100万Token上下文，在通用能力、推理能力、中英文知识处理及Agent能力上有显著提升。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-image-2.0-2026-03-03","description":"阿里通义千问 Qwen-Image 2.0 图像生成与编辑模型，支持中英文文字渲染、自然语言编辑、多图参考和高质量视觉内容创作。该名称固定到 2026 年 3 月 3 日版本，适合需要版本一致性的应用。","tags":"图片","vendor_id":4,"created_time":1773899374,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","image-generation"],"billing_labels":["按次计费"]},{"model_name":"qwen3-30b-a3b","description":"Qwen3-30B-A3B是Qwen3系列的最新成员，采用先进的密集与混合专家（MoE）架构，创新性地实现了思考模式与非思考模式的无缝融合，支持在对话中灵活切换。尽管参数规模更为精简，其推理能力已可媲美QwQ-32B，通用能力显著超越Qwen2.5-14B，在同规模模型中达到业界SOTA水平。凭借在推理、指令遵循、智能代理和多语言支持等方面的全面突破，Qwen3-30B-A3B为高效智能应用提供了理想的解决方案。","tags":"文本","vendor_id":4,"created_time":1762411851,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-audio-3.0-tts-flash","description":"是面向实时交互场景优化的高性能语音合成大模型。相比前一版本，模型支持更多小语种和中文方言，提升了方言发音的正宗程度，并增强了 free-style 指令遵循能力和细粒度标签控制能力，可更灵活地控制情绪、语气、角色、语速、音量等表达方式。同时，模型在噪声、混响等复杂声学条件下具备更强鲁棒性，提升了音质、清晰度和整体表现力。Flash 版本重点优化实时合成体验，适用于语音助手、实时对话、智能客服等低延迟交互场景。","tags":"语音","vendor_id":4,"created_time":1789460053,"quota_type":0,"model_ratio":0.562065,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI文本转语音"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.5-preview","description":"OpenAI GPT-4.5 研究预览模型，侧重广泛知识、自然对话、写作创造力和细腻的意图理解，适合开放式内容生成与高质量交流。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.5,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"Moonshot-Kimi-K2-Instruct","description":"月之暗面开源基础大模型，采用MoE架构，1T总参数，32B激活参数。支持256K上下文，在代码生成、Agent工具调用、数学推理方面表现卓越。","tags":"文本","vendor_id":9,"created_time":1737782400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-sonnet-4-5-20250929-thinking","description":"Claude Sonnet 4.5增强思考版，具备可见的逐步推理能力。展示完整的思考过程，适合需要透明化决策逻辑的复杂任务场景。继承主模型的所有核心能力，在编码、数学推理和代理构建方面表现卓越。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-1.5-pro","description":"Google Gemini 1.5 Pro 高能力多模态模型，擅长长文档、代码、图像、音频和视频理解，适合复杂分析与长上下文任务。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.63,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-image-2-pro","description":"GPT Image 2 Pro 是一套基于 OpenAI 图像生成技术构建的专业化 AI 图像创作工作空间，主要面向设计师、营销人员及电商从业者。该平台通过增强提示词处理、多参考图融合及精准的文字渲染能力，支持 4K 级图像生成、智能编辑与一致性控制，适用于制作产品海报、UI 模型及高保真视觉素材。","tags":"图片","vendor_id":1,"created_time":1776910215,"quota_type":1,"model_ratio":0,"model_price":0.25,"owner_by":"","completion_ratio":0,"enable_groups":["A2","B1","default","A1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"qwen-deep-research","description":"面对复杂研究课题，传统的手动搜索耗时费力，而大模型结合联网搜索的功能也往往难以进行深度、系统的分析。深入研究模型（Qwen-Deep-Research）能自动规划研究步骤、执行多轮的深入搜索与信息整合，并最终生成一份结构化的研究报告。\n\n","tags":"研究","vendor_id":4,"created_time":1769050784,"quota_type":0,"model_ratio":27,"model_price":0,"owner_by":"","completion_ratio":3.0185,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-flash-2025-10-15","description":"轻量级多模态视觉-语言模型2025年10月15日版本。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"wan2.7-image","description":"万相2.7图像生成与编辑模型支持文生图、文生组图、图生组图、图像编辑、多图参考生成，满足多样化生成与集成需求。\n\n","tags":"图片","vendor_id":4,"created_time":1775048557,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gemini-flash-latest-nothinking","description":"Google Gemini Flash 系列高速多模态模型，兼顾推理、长上下文、工具调用和成本效率，适合实时及高并发应用。该版本不启用扩展思考，优先提供更低延迟和更高吞吐量。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["gemini-vertex","C2","G1","G2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5.2-high","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.4-mini-2026-03-17","description":"OpenAI GPT-5.4 系列轻量模型，在代码、推理和工具调用能力与成本之间取得平衡，适合高并发及中等复杂度任务。该名称固定到 2026 年 3 月 17 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.2","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","codex渠道-gpt","default","G2","A2","G1","A1","B1"],"supported_endpoint_types":["gemini","openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-2025-09-23-search-thinking","description":"通义千问旗舰模型2025年9月23日版本联网搜索思考模式。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"wan2.6-image","description":"通义万相2.6-图像生成，全能图像生成模型，支持图文一体化推理生成，具备多图创意融合、商用级一致性、美学要素迁移与镜头光影精确控制，全面提升图像生成的一致性、可控性和表现力。","tags":"图片","vendor_id":4,"created_time":1768987687,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-mini","description":"OpenAI GPT-5 系列轻量模型，兼顾推理能力、速度和成本，适合高并发对话、代码辅助、摘要与结构化数据处理。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro","description":"Google于2025年11月发布的最先进推理模型，是当前最强大的AI模型。具备前所未有的推理深度和细微差别理解能力，在高级推理、复杂指令遵循、工具使用和代理用例方面显著超越Gemini 2.5 Pro。支持1百万token上下文窗口（约1500页文本或30000行代码），世界领先的多模态理解能力（文本、图像、视频、音频），适合处理复杂问题和创意任务。","tags":"文本","vendor_id":6,"created_time":1766546328,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["default","A1","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"cs-gemini-2.5-flash-deepsearch","description":"Google Gemini 2.5 Flash深度搜索版，集成增强的Google搜索能力进行深度信息检索和分析。结合Gemini 2.5 Flash的高速性能和搜索引擎的实时数据，提供更深入、更全面的信息挖掘。适用于需要多源信息整合、深度研究分析的复杂场景。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.4-2026-03-05","description":"OpenAI GPT-5.4 通用旗舰模型，擅长编码、复杂推理、知识工作和工具调用，适合作为专业应用与智能体工作流的核心模型。该名称固定到 2026 年 3 月 5 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1773730249,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-video-3","description":"xAI推出的Grok Imagine视频生成工具（v0.9模型）。支持文本转视频和图像转视频。自动生成背景音乐、音效和环境音频，采用电影级物理模拟。支持现实、动画、抽象艺术等多种风格，提供多种宽高比，完美适配各社交平台。","icon":"XAI.Color","tags":"视频","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"happyhorse-1.0-i2v-720P","description":"首帧图生视频","tags":"视频","vendor_id":4,"created_time":1780043327,"quota_type":1,"model_ratio":0,"model_price":0.9,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"qwen-image-2.0","description":"阿里通义千问 Qwen-Image 2.0 图像生成与编辑模型，支持中英文文字渲染、自然语言编辑、多图参考和高质量视觉内容创作。","tags":"图片","vendor_id":4,"created_time":1773899330,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","image-generation"],"billing_labels":["按次计费"]},{"model_name":"claude-opus-4-7","description":"Anthropic Claude Opus 4.7 旗舰模型，面向高难度软件工程、复杂推理和长周期自主智能体任务，强调深度分析与可靠执行。","tags":"文本","vendor_id":2,"created_time":1776388571,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","C2","Claude-Aws","default","G1","A2","Claude Max","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-5","description":"Anthropic 于 2026 年 6 月 30 日发布的新一代 Sonnet 模型，定位是“速度与智能的最佳平衡”。它相比 Claude Sonnet 4.6 在代码开发、工具调用、多步骤代理任务、推理和知识工作上有明显提升，尤其适合软件工程、自动化流程、浏览器/终端工具使用和长任务执行。","tags":"文本","vendor_id":2,"created_time":1782879946,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["C2","G1","A2","Claude Max","Claude-Aws","default","G2","A1","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash-search","description":"Google Gemini 2.5 Flash搜索增强版，集成Google搜索能力进行实时信息检索。结合Gemini 2.5 Flash的高速性能和搜索引擎的实时数据，提供最新、最准确的信息。适用于需要实时信息更新的场景，如新闻分析、市场研究、事实查证等任务。","tags":"文本","vendor_id":6,"created_time":1766666064,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.34,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"veo_3_1-fast","description":"Google Veo 3.1快速版本（下划线命名格式）。优化生成速度，保持Veo 3.1的核心改进功能，适合需要快速响应的视频生成任务。","tags":"视频","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-flash-preview-thinking-*","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为预览版本，能力和行为可能随服务商更新而调整。名称保留渠道提供的字面量 * 后缀，用于其自身的路由约定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5-nano-2025-08-07","description":"OpenAI GPT-5 系列最轻量模型，针对低延迟和大规模调用优化，适合分类、抽取、排序和简单生成任务。该名称固定到 2025 年 8 月 7 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.025,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-image-2.5-sunburst","description":"OpenAI 于 2026 年 9 月推出的旗舰级图像生成与编辑模型。该模型主打高画质与精准编辑控制，具备更自然的光影质感及出色的主体一致性保持能力。它主要面向对细节要求严苛的商业视觉、广告创意与高精产品图渲染等专业工作流。","tags":"图片","vendor_id":1,"created_time":1788922101,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["A1","B1","codex渠道-gpt","default"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI生图","OpenAI改图"],"billing_labels":["按次计费"]},{"model_name":"o1","description":"OpenAI推理型模型系列首次引入，核心特性是在回复前花费更多时间思考。在数学竞赛（AIME）达到83%正确率，相当于全美前500名学生水平；博士级科学问题（GPQA）准确率78%，超过人类专家69.7%；编程竞赛超过83%的专业选手。支持128k上下文，基于大规模强化学习在思维链上训练。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-o3-pro","description":"OpenAI史上最强AI模型（2025年6月发布），在数学基准AIME 2024中超越Google Gemini 2.5 Pro，在博士级科学测试GPQA Diamond中击败Anthropic Claude 4 Opus。可访问所有工具：网络搜索、文件分析、Python使用、视觉推理等。o3整体降价80%，提供卓越性价比。当前名称为按次计费别名，实际模型名称为 o3-pro。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-flash-preview-c","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。当前名称为按次计费别名，实际模型名称为 gemini-3-flash-preview。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gemini-pro-latest","description":"Google Gemini Pro 系列高能力多模态模型，面向复杂推理、代码、长上下文分析和专业知识任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"qwen-plus-search-thinking","description":"通义千问Plus联网搜索思考模式，结合实时联网搜索和深度推理能力处理复杂问题。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":13.333333333333,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-thinking","description":"Google Gemini 3.1 系列增强推理路由模型，面向复杂分析、代码规划和多步骤智能体任务，具体底层版本由渠道配置决定。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"qwen3-vl-flash-2025-10-15-search","description":"轻量级多模态视觉-语言模型2025年10月15日版本联网搜索。","tags":"联网","vendor_id":4,"created_time":1769691324,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-preview-thinking-c","description":"Google Gemini 3 Pro预览版思考配置，启用可见的推理过程展示。通过thought signatures维护多步骤工作流中的推理链，展示完整的思考过程。帮助用户理解模型如何处理复杂问题，适合需要透明化决策逻辑的教学、研究和调试场景。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"o4-mini-dr","description":"OpenAI o4-mini深度研究（Deep Research）版本，结合o4-mini的快速响应和深度研究能力。提供经济的深度分析方案，适合需要快速完成研究任务的场景。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v4-pro","description":"DeepSeek-V4-Pro 是 DeepSeek 于 2026 年 4 月发布的旗舰级混合专家模型（MoE），拥有 1.6 万亿总参数及 490 亿激活参数。该模型支持 100 万 token 的超长上下文窗口，专为复杂推理、编程及长任务智能体场景设计，在数学、逻辑与工程任务中展现出卓越的性能。","tags":"文本","vendor_id":3,"created_time":1778230072,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["A1","A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"(tier(\"base\", p * 4.5 + c * 13.5 + cr * 0.15)) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 9 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 12 ? 2 : 1) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 14 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 18 ? 2 : 1)"},{"model_name":"qwen3-vl-plus-2025-12-19-search","description":"多模态视觉-语言模型2025年12月19日版本联网搜索。","tags":"联网","vendor_id":4,"created_time":1769691429,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-gpt-4.1","description":"OpenAI GPT-4.1 旗舰非推理模型，针对编码、指令遵循和长上下文理解优化，适合软件开发、文档分析和工具调用。当前名称为按次计费别名，实际模型名称为 gpt-4.1。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.025,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"net-gpt-4o","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。当前名称为服务侧路由别名，底层模型为 gpt-4o。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.3-codex-xhigh","description":"OpenAI GPT-5.3 系列智能体编码模型，面向复杂软件工程、跨文件修改、长时间任务、终端操作和代码审查优化，适合自主开发工作流。该条目将推理强度设为超高档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":1,"created_time":1770607044,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-6","description":"Anthropic Claude Sonnet 4.6 高性能均衡模型，面向复杂编码、长程智能体、工具使用和专业知识工作，在速度与能力之间取得平衡。","tags":"文本","vendor_id":2,"created_time":1771744572,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A2","Claude Max","Claude-Aws","default","G2","B1","C2","G1","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-preview-thinking","description":"Google Gemini 3 Pro预览版思考模式，启用可见的推理过程展示。通过thought signatures维护多步骤工作流中的推理链，展示完整的思考过程。帮助用户理解模型如何处理复杂问题，适合需要透明化决策逻辑的教学、研究和调试场景。","tags":"文本","vendor_id":6,"created_time":1766666232,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G1","G2","gemini-vertex","A1","A2","B1","C2","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.5-openai-compact","description":"OpenAI GPT-5.5 的上下文压缩路由，用于通过 Responses Compact 接口压缩长对话状态，降低后续请求的上下文负担。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["default","G1","G2","A1","B1","C2","codex渠道-gpt"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5) : tier(\"long_context\", p * 10 + c * 45 + cr * 1)"},{"model_name":"o1-mini","description":"OpenAI o1系列速度更快、成本更低的精简版本，擅长编码、数学和科学任务。最大输出限制65,536 tokens（是preview的两倍），成本仅3美元/百万输入tokens。专注于STEM领域，不需要广泛通用知识，适合追求成本效益的编程任务、数学问题和科学推理。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-haiku-4-5","description":"Anthropic Claude Haiku 4.5 高速高性价比模型，强化编码、智能体和工具调用能力，适合实时交互与大规模生产工作负载。","tags":"文本","vendor_id":2,"created_time":1785210665,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","A2","Claude-Aws","G1","G2","B1","C2","Claude Max","default"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-6-thinking","description":"Anthropic Claude Sonnet 4.6 高性能均衡模型，面向复杂编码、长程智能体、工具使用和专业知识工作，在速度与能力之间取得平衡。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1771744600,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["B1","default","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-high","description":"Google Gemini 3 Pro高推理强度版，提供最深度的思考和分析能力。通过thinking_level参数控制推理深度，适合需要最高质量输出的复杂任务。在高级推理、复杂指令遵循和多步骤推理链方面表现卓越。支持1百万token上下文窗口，适合科研分析、战略规划、复杂编码等需要深度思考的场景。","tags":"文本","vendor_id":6,"created_time":1766666204,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.0-flash-thinking-exp","description":"Google Gemini 2.0 Flash 高速多模态模型，支持长上下文、工具调用、结构化输出以及文本、图像、音频和视频理解。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为实验版本，主要用于提前验证新能力。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"grok-3-deepsearch","description":"xAI Grok 3 通用大语言模型，擅长推理、代码、实时信息分析和工具使用，适合对话、研究与复杂问题求解。该版本面向深度搜索与资料综合，适合研究型问答。","icon":"XAI.Color","tags":"研究","vendor_id":8,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-3.5-turbo-16k","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G1","G2","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-preview-search-thinking","description":"通义千问旗舰模型预览版联网搜索思考模式。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-thinking","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":4,"created_time":1772075178,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-pro-preview-high","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":6,"created_time":1771912449,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-5-20250929","description":"Anthropic于2025年9月发布的旗舰级AI模型，被誉为\"全球最佳编码模型\"。在SWE-bench Verified上达到77.2%的基准准确率（使用并行测试时计算可达82%）。支持200K tokens上下文窗口（beta模式可扩展至1M tokens），输出可达64K tokens。在计算机使用能力方面领先，OSWorld基准测试得分61.4%。擅长复杂代理构建、浏览器任务处理、数学推理和多步骤编码任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1762411850,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A2","C2","Claude-Aws","G1","B1","Claude Max","default","G2","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-flash","description":"轻量级多模态视觉-语言模型，侧重快速推理。支持文本、图像和视频输入，262K超长上下文，响应速度更快。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-thinking","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-pro-preview-c","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务。当前名称为按次计费别名，实际模型名称为 gemini-3.1-pro-preview。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-4-all","description":"OpenAI GPT-4 系列高能力通用模型，擅长复杂指令遵循、逻辑推理、代码生成和专业文本处理，是早期 GPT-4 代旗舰模型。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"suno_music_open","description":"Suno AI音乐生成模型的开放服务版本，核心能力与suno_music一致，可根据歌词、主题或音乐风格生成包含人声与器乐的完整歌曲，支持多语言和多种音乐类型，适合个人创作、音乐教育、商业制作与游戏内容开发。","tags":"音乐","vendor_id":1,"created_time":1785151238,"quota_type":1,"model_ratio":0,"model_price":0.35,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","Suno Open提交","Suno Open任务查询","Suno Open任务详情"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.8-flash","description":"Gemini 3.8 Flash 是 Google 于 2026 年 9 月发布的轻量级高性能模型，专为长周期软件工程、自主智能体及复杂企业级工作流优化。该模型在保持 Flash 系列高效与低成本优势的同时，显著提升了多步推理与代码处理能力，并同步推出了针对网络安全领域的专用版本。","tags":"文本","vendor_id":6,"created_time":1788584142,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["G1","gemini-vertex","A1","G2","A2","B1","C2","default"],"supported_endpoint_types":["gemini","openai","Gemini"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-flash-2026-07-15","description":"Qwen3.7-Flash-2026-07-15 是 Qwen3.7 原生视觉语言 Flash 快照，强化图像、视频理解、真实世界感知、Search Agent、CI Agent 和多模态 Coding，支持约 100 万 token 上下文，适合高并发多模态智能体任务。","tags":"文本","vendor_id":4,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 32000\n  ? tier(\"standard\", p * 0.2 + c * 0.8 + cr * 0.04)\n  : (len \u003c= 256000\n    ? tier(\"medium_context\", p * 0.6 + c * 2.4 + cr * 0.12)\n    : tier(\"long_context\", p * 1.2 + c * 4.8 + cr * 0.24))"},{"model_name":"net-o1-preview","description":"OpenAI o1系列早期预览版本，设计用于利用广泛通用知识解决复杂问题。支持更广泛的STEM以外领域的世界知识，最大输出限制32,768 tokens。适合法律分析、科研推理、金融决策等需要高精度逻辑和博士级推理能力的场景。成本15美元/百万输入tokens。当前名称为服务侧路由别名，底层模型为 o1-preview。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-flash","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","A2","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.6-flash","description":"Gemini 3.6 Flash 是谷歌于 2026 年 7 月发布的最新高效能 AI 模型，专为开发者及智能体（Agentic）应用场景优化。该模型在保持极高运行速度与低成本的同时，显著提升了代码生成与复杂逻辑推理能力，尤其在处理长周期、多步骤的自动化任务时表现出色。","tags":"文本","vendor_id":6,"created_time":1785053906,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["A1","C2","default","G1","gemini-vertex","A2","B1","G2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-codex-mini","description":"OpenAI GPT-5.1-Codex轻量版本，专注于快速代码生成和补全。在保持良好代码质量的同时提供更快的响应速度和更低的成本，适合日常编码辅助、代码补全和快速原型开发。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1763611674,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash-image","description":"Google DeepMind于2025年8月推出的AI图像生成与编辑模型（代号Nano Banana），具备良好的性价比。支持文本驱动图像生成、多图混合、多轮编辑、风格迁移、角色一致性保持、局部编辑等功能。每个提示最多3张输入图片和10张输出图片，支持10种宽高比，所有生成内容带有水印和SynthID数字标记。","icon":"Gemini.Color","tags":"图片","vendor_id":6,"created_time":1762411851,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["default"],"supported_endpoint_types":["gemini","openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-0125-preview","description":"OpenAI GPT-4 Turbo 模型，在保持较强推理与代码能力的同时提供更长上下文、更高速度和更低成本，支持文本与视觉任务。该名称固定到 2024 年 1 月 25 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o1-preview-all","description":"OpenAI o1-preview全能版本，提供广泛通用知识的完整推理能力。适合需要跨领域知识和博士级推理的综合性复杂问题解决。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-mini-480p-i2v","description":"支持 480p 分辨率。Seedance 2.0 Mini 轻量视频生成模型，仅支持文生视频和图生视频，面向成本敏感、高并发和快速内容生产场景。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.125,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-codex-low","description":"OpenAI GPT-5-Codex低资源版本，优先考虑响应速度和成本效益。适合简单代码补全、快速原型和日常编码辅助，提供经济实惠的代码生成方案。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-gemini-3-pro-preview","description":"Google Gemini 3 Pro预览版，提供早期访问体验。最先进的推理模型，具备前所未有的推理深度和细微差别理解能力。在高级推理、复杂指令遵循、工具使用和代理用例方面显著超越Gemini 2.5 Pro。支持1百万token上下文窗口，世界领先的多模态理解能力。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.05,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按次计费"]},{"model_name":"grok-3","description":"xAI Grok 3 通用大语言模型，擅长推理、代码、实时信息分析和工具使用，适合对话、研究与复杂问题求解。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1762411851,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"kimi-k2.6","description":"Kimi K2.6是月之暗面（Moonshot AI）于2026年4月发布并开源的生产级原生多模态智能体编程模型，基于1万亿总参数、320亿激活参数的MoE架构（384个专家，MLA注意力机制）打造，具备256K上下文窗口与原生文本/图像/视频理解能力（搭载MoonViT视觉编码器），核心聚焦于长周期编码、主动自主执行与多智能体集群编排，支持300个子智能体并行协同及4000+步协调、12小时以上连续自主运行，在SWE-Bench Pro（58.6%）、Terminal-Bench 2.0（66.7%）、HLE w/ tools（54.0%）等基准上取得开源SOTA并部分反超闭源顶尖模型。","tags":"文本","vendor_id":7,"created_time":1784110303,"quota_type":0,"model_ratio":3.25,"model_price":0,"owner_by":"","completion_ratio":4.153846153846,"cache_ratio":0.169231,"enable_groups":["A2","B1","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"cs-grok-4-1-thinking-1129","description":"xAI Grok 4.1 高能力通用模型，强化推理、代码、工具调用和智能体任务，在复杂工作流中兼顾准确性与执行效率。当前名称为按次计费别名，实际模型名称为 grok-4-1-thinking-1129。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为指定发布批次的固定版本，便于保持行为一致性。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.07,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"o1-mini-2024-09-12","description":"OpenAI o1-mini模型2024年9月12日初始发布版本，提供稳定的STEM领域推理能力。适合需要版本一致性的编程任务、数学问题和科学推理应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-image-2.5-flare","description":"OpenAI 于 2026 年 9 月推出的 API 默认图像生成模型。它相比前代降低了约 50% 的生成延迟，兼顾更快的生成速度与出色的多轮精准编辑能力，主要面向社交内容创作、产品原型设计及高频批量出图等应用场景。","tags":"图片","vendor_id":1,"created_time":1788922060,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["codex渠道-gpt","default","A1","B1"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI生图","OpenAI改图"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.0-flash-lite-preview","description":"Google Gemini 2.0 Flash-Lite 轻量多模态模型，面向低延迟、低成本和高吞吐任务，支持文本、图像、音频与视频输入。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"o1-all","description":"OpenAI o1全能版本，整合o1系列所有能力的综合模型。提供标准推理性能，适合需要全面o1能力但不需要特定优化的通用推理任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.5,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"o3-mini-high-all","description":"OpenAI o3-mini-high全能版本，结合高推理强度和全面功能。提供最深入的思考和最高准确度，适合需要精确推理的复杂任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"grok-4.20-0309-non-reasoning","description":"grok-4.20-0309-non-reasoning 是 xAI 于 2026 年 3 月 9 日发布的 Grok 4.20 系列的非推理（Non-Reasoning）高速变体，它在保留与推理版相同底层能力与 100 万 token 上下文窗口的前提下，刻意跳过扩展链式思维（CoT）步骤，专注于提供极低延迟的直接响应，适用于高吞吐量、对实时性要求高于深层逻辑推理的场景（如客服、批量处理、工具调用循环）。该模型支持文本与图像（JPEG/PNG 等）的多模态输入、原生函数调用（Function Calling）、结构化输出及文档理解，具备业界领先的低幻觉率与严格的提示词遵循能力，API 定价约为输入 1.25/百万 token（缓存 0.20）、输出 $2.50/百万 token，在牺牲部分复杂多步推理精度的同时换取了约一半的首 token 延迟与更高的吞吐表现。","tags":"文本","vendor_id":8,"created_time":1784110011,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.16,"create_cache_ratio":0.16,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-7-thinking-c","description":"Anthropic Claude Opus 4.7 旗舰模型，面向高难度软件工程、复杂推理和长周期自主智能体任务，强调深度分析与可靠执行。当前名称为按次计费别名，实际模型名称为 claude-opus-4-7-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1776929900,"quota_type":1,"model_ratio":0,"model_price":0.18,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-1106-preview","description":"OpenAI GPT-4 Turbo 模型，在保持较强推理与代码能力的同时提供更长上下文、更高速度和更低成本，支持文本与视觉任务。该名称固定到 2023 年 11 月 6 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o1-preview-2024-09-12","description":"OpenAI o1-preview模型2024年9月12日初始发布版本，提供广泛通用知识的复杂问题解决能力。适合需要稳定博士级推理能力的法律分析、科研推理和金融决策等场景。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.5-flash-lite","description":"Google Gemini 3.5 Flash-Lite 轻量高速多模态模型，面向成本敏感、高并发和低延迟场景，适合分类、抽取与实时交互。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.3333,"enable_groups":["A2","B1","C2","default","G1","G2","A1","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1-nano-2025-04-14","description":"OpenAI GPT-4.1 系列最轻量模型，提供低延迟和低成本的长上下文处理，适合分类、抽取、自动补全及高吞吐任务。该名称固定到 2025 年 4 月 14 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-coder-flash","description":"\"甜品级\"编程模型，305亿参数，其中33亿参数被激活。卓越的Agent能力，原生支持256K tokens，可扩展至1M tokens。兼顾性能与成本效率。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-coder-plus","description":"通义千问编程模型顶配版本，接近Qwen3-Coder-480B及Claude Sonnet-4、GPT4.1等领先闭源模型。支持128K上下文，具备卓越的编程能力和Agent特性。","tags":"文本","vendor_id":4,"created_time":1762411851,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-flash-lite-latest-thinking","description":"Google Gemini Flash-Lite 轻量多模态模型，强调低延迟、低成本和大规模吞吐，适合分类、抽取与快速对话。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-3.5-turbo-0613","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。该名称固定到 2023 年 6 月 13 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-sonnet-4-20250514-thinking","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-20250514-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.5-instant","description":"OpenAI GPT-5.5 前沿通用模型，面向复杂编码、研究分析和真实业务任务，强化长链路推理、工具协作与高质量专业输出。该版本强调即时响应和低延迟，适合高频交互场景。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G2","C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-2026-02-15-search","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。该名称固定到 2026 年 2 月 15 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1772075117,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-audio-3.0-tts-plus","description":"是面向高质量语音生成场景打造的高性能语音合成大模型。相比前一版本，模型支持更多小语种和中文方言，显著提升方言发音的正宗程度，并增强了 free-style 指令遵循能力和细粒度标签控制能力，可更准确地控制情绪、语气、角色、语速、音量和合成风格。同时，模型在噪声、混响等复杂声学条件下具备更强鲁棒性，进一步提升了音质、清晰度、分辨率和整体表现力。Plus 版本更强调合成效果和细节表现，适用于有更高音质、自然度和表现力要求的专业场景，如内容创作、有声书、影视配音、品牌声音设计和高品质语音服务。","tags":"语音","vendor_id":4,"created_time":1789460106,"quota_type":0,"model_ratio":0.74942,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI文本转语音"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro-preview-05-06","description":"Google Gemini 2.5 Pro预览版（2025年5月6日版本），提供早期访问体验。具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。支持原生音频输出、高级安全防护和Project Mariner的计算机使用功能。配备\"Deep Think\"实验性增强推理模式，专门针对复杂数学和编码问题进行优化。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-sol-openai-compact","description":"GPT-5.6 Sol 的 OpenAI Responses Compact 路由别名，面向质量优先的复杂推理与代码工作流，用于压缩长对话状态，实际底层模型由渠道配置决定。","tags":"文本","vendor_id":1,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G1","G2","A1","B1","C2","codex渠道-gpt","default"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5 + cc * 6.25) : tier(\"long_context\", p * 10 + c * 45 + cr * 1 + cc * 12.5)"},{"model_name":"o3-mini-2025-01-31","description":"OpenAI o3-mini模型2025年1月31日正式发布版本，首次向ChatGPT免费用户开放推理模型的里程碑版本。提供稳定的快速推理能力和三级推理强度调节。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-2025-04-28-search","description":"通义千问Turbo 2025年4月28日版本联网搜索。","tags":"联网","vendor_id":4,"created_time":1769691158,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"wan2.7-t2v-720P","description":"阿里通义万相 Wan2.7 文生视频模型，根据文本提示生成具备连贯运动与镜头语言的视频内容，适合短片与商业创作。","tags":"视频","vendor_id":4,"created_time":1775802130,"quota_type":1,"model_ratio":0,"model_price":0.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-3.5-turbo-instruct","description":"OpenAI 基于 GPT-3.5 系列推出的指令补全模型，面向单轮文本生成、改写、摘要和传统 Completions 工作流优化，响应速度快、调用成本较低。","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-chat-2025-11-13","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该名称固定到 2025 年 11 月 13 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-preview-c","description":"Google Gemini 3 Pro预览版标准配置，提供早期访问体验。最先进的推理模型，具备前所未有的推理深度和细微差别理解能力。在高级推理、复杂指令遵循、工具使用和代理用例方面显著超越Gemini 2.5 Pro。支持1百万token上下文窗口，世界领先的多模态理解能力。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["default","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-pro-thinking-*","description":"Google Gemini 2.5 Pro 旗舰推理多模态模型，擅长复杂代码、数学、长文档和视频理解，支持长上下文与工具调用。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。名称保留渠道提供的字面量 * 后缀，用于其自身的路由约定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"grok-4-fast-non-reasoning","description":"xAI Grok 4 旗舰通用模型，面向复杂数学、科学、代码和智能体任务，支持长上下文、工具调用与实时信息处理。该版本不启用扩展思考，优先提供更低延迟和更高吞吐量。该版本针对响应速度和吞吐量进行了优化。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1762411851,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-1-20250805-thinking","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 8 月 5 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":2,"created_time":1766650225,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash","description":"Google于2025年6月正式发布的高速AI模型，专为速度、效率和规模化而优化，提供最佳性价比。适用于大规模摘要生成、响应式聊天应用和数据提取等高吞吐量企业任务。支持多模态输入（文本、代码、图像、音频、视频），最大输入1,048,576 tokens，输出65,535 tokens。具备思考能力、Google搜索整合、函数调用、结构化输出、上下文缓存等高级功能。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.33333,"enable_groups":["A1","A2","C2","default","G1","gemini-vertex","B1","G2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-search-thinking","description":"通义千问Turbo联网搜索思考模式，快速响应结合实时联网和推理能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-image-preview-2k","description":"Google Gemini 3 Pro Image预览版支持打印级质量输出。具备思考推理能力和搜索Grounding功能，可生成动态内容。支持12+种语言的多语言文字渲染，高级图像混合可融合至14张图像，专注于复杂、多轮、高保真的专业级生成任务。","icon":"Gemini.Color","tags":"图片","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-edits","image-generation"],"billing_labels":["按次计费"]},{"model_name":"suno_persona_create","description":"Suno AI Personas人设创建功能（V4新增），捕捉并保存曲目的人声、风格和氛围，创建独特的音乐人设。可将特定的音乐风格、声线特征延续到后续创作中，保持品牌一致性。适合系列音乐创作、艺术家风格模仿、音乐品牌建设。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v3.2-search-thinking","description":"DeepSeek V3.2联网搜索思考模式，结合联网搜索和深度推理能力。","tags":"联网","vendor_id":3,"created_time":1737782400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-5-all","description":"Claude Sonnet 4.5 是 Anthropic 于 2025 年 9 月发布的高性能推理模型，主要面向复杂的长链路任务、智能体构建及专业代码编程场景。该模型支持 20 万上下文窗口及多模态视觉输入，曾以卓越的逻辑推理和工具使用能力在编程基准测试中表现突出。随着技术迭代，目前该模型已由更新的版本（如 Sonnet 4.6 及后续系列）接替。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"net-gpt-3.5-turbo","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。当前名称为服务侧路由别名，底层模型为 gpt-3.5-turbo。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.002,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v3.2","description":"DeepSeek V3.2最新版本，代表推动开源大语言模型发展的最新进展，继续推进DeepSeek模型系列的能力边界。","tags":"文本","vendor_id":3,"created_time":1766546328,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["G2","A1","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-codex-medium","description":"OpenAI GPT-5-Codex中等性能版本，在代码质量、响应速度和成本之间提供最佳平衡。适合大多数日常编程任务和中等复杂度的代码开发。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-minimal","description":"OpenAI GPT-5 系列轻量模型，兼顾推理能力、速度和成本，适合高并发对话、代码辅助、摘要与结构化数据处理。该条目将推理强度设为最小档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"kimi-k2.5","description":"Kimi-K2.5 是月之暗面（Moonshot AI）于 2026 年 1 月 27 日正式发布并开源的新一代全能型开源大模型，被定义为 Kimi 系列迄今最智能、综合能力最强的模型。","tags":"文本","vendor_id":9,"created_time":1769656763,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5.25,"enable_groups":["default","G1","G2","A1","A2","B1","C2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-plus-latest","description":"通义千问Plus最新版本，效果、速度、成本均衡，支持超长100万Token上下文和双推理模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-thinking","description":"通义千问旗舰模型思考模式，用于复杂问题的深度推理分析。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":3.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-sonnet-4-5-20250929","description":"Anthropic于2025年9月发布的旗舰级AI模型，被誉为\"全球最佳编码模型\"。在SWE-bench Verified上达到77.2%的基准准确率（使用并行测试时计算可达82%）。支持200K tokens上下文窗口（beta模式可扩展至1M tokens），输出可达64K tokens。在计算机使用能力方面领先，OSWorld基准测试得分61.4%。擅长复杂代理构建、浏览器任务处理、数学推理和多步骤编码任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-image-2k","description":"Google Gemini 3 Pro Image 2K版本，支持2K分辨率图像生成。基于推理优先架构，具备搜索接地功能，可获取实时准确信息。支持多轮对话式编辑，精准处理复杂指令和文字渲染，适合专业级图像生成任务。","tags":"图片","vendor_id":6,"created_time":1766546663,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"qwen3-max-2025-09-23-thinking","description":"通义千问旗舰模型2025年9月23日版本思考模式。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-2026-02-15","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该名称固定到 2026 年 2 月 15 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1771925901,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"net-o1-mini","description":"OpenAI o1系列速度更快、成本更低的精简版本，擅长编码、数学和科学任务。最大输出限制65,536 tokens（是preview的两倍），成本仅3美元/百万输入tokens。专注于STEM领域，不需要广泛通用知识，适合追求成本效益的编程任务、数学问题和科学推理。当前名称为服务侧路由别名，底层模型为 o1-mini。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"grok-imagine-image-quality","description":"Grok Imagine 是 xAI 推出的图像生成系列模型，其中 grok-imagine-image-quality 是其 API 中用于高质量文生图任务的特定模型标识。该系列基于 xAI 自研的 Aurora 架构，采用自回归混合专家（MoE）技术，在 photorealism（写实感）、复杂场景构图及文字渲染方面表现出色，主要面向创意设计、营销内容及视觉创作等应用场景。","tags":"图片","vendor_id":8,"created_time":1786682756,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI生图","OpenAI改图"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-720p-i2v","description":"支持 720p 分辨率。Seedance 2.0 标准视频生成模型，仅支持文生视频和图生视频，适合多镜头叙事、电影感镜头与音画同步创作。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.9,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"wan2.7-i2v-720P","description":"阿里通义万相 Wan2.7 图生视频模型，根据首帧图像与文本提示生成连贯视频，强化主体一致性、运动表现和镜头控制。","tags":"视频","vendor_id":4,"created_time":1775801973,"quota_type":1,"model_ratio":0,"model_price":0.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-4.5-preview-2025-02-27","description":"OpenAI GPT-4.5 研究预览模型，侧重广泛知识、自然对话、写作创造力和细腻的意图理解，适合开放式内容生成与高质量交流。该名称固定到 2025 年 2 月 27 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1766658117,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.5,"enable_groups":["C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-search","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"联网","vendor_id":4,"created_time":1775140319,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-7-c","description":"Anthropic Claude Opus 4.7 旗舰模型，面向高难度软件工程、复杂推理和长周期自主智能体任务，强调深度分析与可靠执行。当前名称为按次计费别名，实际模型名称为 claude-opus-4-7。","tags":"文本","vendor_id":2,"created_time":1776929913,"quota_type":1,"model_ratio":0,"model_price":0.18,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-thinking-all","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwq-plus-2025-03-05","description":"QWQ-Plus 2025年3月5日版本，特定日期快照版本。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.8,"model_price":0,"owner_by":"","completion_ratio":2.5,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-o3-pro-all","description":"OpenAI史上最强AI模型（2025年6月发布），在数学基准AIME 2024中超越Google Gemini 2.5 Pro，在博士级科学测试GPQA Diamond中击败Anthropic Claude 4 Opus。可访问所有工具：网络搜索、文件分析、Python使用、视觉推理等。o3整体降价80%，提供卓越性价比。当前名称为按次计费别名，实际模型名称为 o3-pro-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai-response"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-mini-720p-i2v","description":"支持 720p 分辨率。Seedance 2.0 Mini 视频生成模型，仅支持文生视频和图生视频，适合多镜头叙事、电影感镜头与音画同步创作。","tags":"视频","vendor_id":5,"created_time":1789723988,"quota_type":1,"model_ratio":0,"model_price":0.25,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["default","B1"],"supported_endpoint_types":["openai","OpenAI视频"],"billing_labels":["按次计费"]},{"model_name":"qwen3-vl-plus-thinking","description":"多模态视觉-语言模型思考模式，支持对视觉内容的深度推理分析。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-video-3-max","description":"支持15s视频，映射grok-video-1.5模型","tags":"视频","vendor_id":8,"created_time":1772352452,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"suno_act_wav","description":"Suno AI音频格式转换功能，将生成的音乐转换为WAV无损音频格式。WAV格式提供最高音质，适合专业音乐制作、后期处理、商业使用和高保真音频存档。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"wan2.7-r2v-1080P","description":"阿里通义万相 Wan2.7 参考生视频模型，可结合参考素材与文本生成视频，保持角色、主体或风格的一致性。","tags":"视频","vendor_id":4,"created_time":1775802148,"quota_type":1,"model_ratio":0,"model_price":1,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.1-chat-latest","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G1","G2","A1","B1","C2","default"],"supported_endpoint_types":["gemini","openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"wan2.7-r2v-720P","description":"阿里通义万相 Wan2.7 参考生视频模型，可结合参考素材与文本生成视频，保持角色、主体或风格的一致性。","tags":"视频","vendor_id":4,"created_time":1775802054,"quota_type":1,"model_ratio":0,"model_price":0.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.3-chat-2026-03-03","description":"OpenAI GPT-5.3 对话模型，面向自然交互、复杂指令、代码辅助和工具调用优化，适合生产级聊天与智能体应用。该名称固定到 2026 年 3 月 3 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-1-20250805","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。该名称固定到 2025 年 8 月 5 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1762411850,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A2","B1","default","A1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"grok-imagine-video-1.5-preview","description":"xAI 2026 年 6 月推出的预览版纯图生视频（I2V）自动画音一体化模型，基于自研 Aurora 自回归 MoE 引擎，图生视频综合榜单排名第一，仅支持图片作为生成基底，单次推理同步输出匹配画面的音效、人声、背景音乐，人物 / 产品主体稳定性、物理运动真实度、音画同步表现大幅领先前代，主打短视频创意快速量产、广告分镜、商品动态短片。","tags":"视频","vendor_id":8,"created_time":1782025086,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"o1-mini-all","description":"OpenAI o1-mini全能版本，提供o1-mini的快速推理能力和全面功能支持。适合需要经济高效的STEM推理和通用编程任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-fast-720p-i2v","description":"支持 720p 分辨率。Seedance 2.0 Fast 高速视频生成模型，仅支持文生视频和图生视频，面向快速预览、高频迭代和低延迟创作。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.56,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gemini-pro-latest-thinking","description":"Google Gemini Pro 系列高能力多模态模型，面向复杂推理、代码、长上下文分析和专业知识任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5.3-chat","description":"OpenAI GPT-5.3 对话模型，面向自然交互、复杂指令、代码辅助和工具调用优化，适合生产级聊天与智能体应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G1","G2","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-luna","description":"GPT-5.6 Luna 是 GPT-5.6 系列的高吞吐、低延迟配置，适合高频对话、批量内容处理、代码辅助和实时智能体工作流。","tags":"文本","vendor_id":1,"created_time":1786443762,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","codex渠道-gpt","G1","A2","C2","default","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-6.1-sol","description":"OpenAI 于**2026 年 9 月 29 日，在 OpenAI DevDay 2026 开发者大会**正式发布的大模型，为 9 月 22 日上线的 GPT-6 Sol 迭代升级版，主打旗舰级任务能力、中端商用定价，面向企业复杂专业任务、AI 智能体开发、软件工程、文档深度解析与多步骤自动化工作流场景，是兼顾性能与成本的主力推理模型OpenAI。","tags":"文本","vendor_id":1,"created_time":1790737185,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["A1","B1","C2","codex渠道-gpt","default","G1","G2"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI对话（响应式）"],"billing_labels":["按量计费"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"standard\", p * 2 + c * 10 + cr * 0.1 + cc * 2.5) : tier(\"long_context\", p * 4 + c * 15 + cr * 0.2 + cc * 5)"},{"model_name":"qwen-plus","description":"效果、速度、成本均衡的通义千问模型，适合中等复杂任务。支持超长100万Token上下文，属于Qwen3系列，支持\"思考模式\"和\"非思考模式\"双推理。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-2026-04-02-search-thinking","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。该名称固定到 2026 年 4 月 2 日版本，适合需要版本一致性的应用。","tags":"联网","vendor_id":4,"created_time":1775140209,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro-nothinking","description":"Google Gemini 2.5 Pro标准版，关闭思考过程显示。旗舰级AI模型，具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。支持原生音频输出、高级安全防护和Project Mariner的计算机使用功能。最大输入1,048,576 tokens，输出65,535 tokens。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["gemini-vertex","B1","C2","default","G1","G2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"seedance-2-0-480p-i2v","description":"支持 480p 分辨率。Seedance 2.0 标准视频生成模型，仅支持文生视频和图生视频，适合多镜头叙事、电影感镜头与音画同步创作。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.45,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"cs-claude-opus-4-1-all","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。当前名称为按次计费别名，实际模型名称为 claude-opus-4-1-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-high-c","description":"Google Gemini 3 Pro高推理强度版，提供最深度的思考和分析能力。通过thinking_level参数控制推理深度，适合需要最高质量输出的复杂任务。在高级推理、复杂指令遵循和多步骤推理链方面表现卓越。支持1百万token上下文窗口，适合科研分析、战略规划、复杂编码等需要深度思考的场景。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.1-codex","description":"OpenAI GPT-5.1系列代码生成模型升级版，在GPT-5-Codex基础上进一步优化。提供更好的代码理解和生成能力，支持多种编程语言和框架，适合专业开发者的复杂编码任务。","tags":"文本","vendor_id":1,"created_time":1763611674,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6-20260203","description":"Anthropic Claude Opus 4.6 旗舰模型，面向复杂软件工程、深度推理、长程智能体和高价值专业工作，强调准确性与持续任务执行。该名称固定到 2026 年 2 月 3 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":2,"created_time":1785210665,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1-mini-2025-04-14","description":"OpenAI GPT-4.1 系列轻量模型，具备优秀的指令遵循、代码能力和长上下文理解，适合兼顾性能、速度与成本的应用。该名称固定到 2025 年 4 月 14 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。","tags":"文本","vendor_id":4,"created_time":1771925887,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。","tags":"文本","vendor_id":4,"created_time":1775140185,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"wan2.7-i2v-1080P","description":"阿里通义万相 Wan2.7 图生视频模型，根据首帧图像与文本提示生成连贯视频，强化主体一致性、运动表现和镜头控制。","tags":"视频","vendor_id":4,"created_time":1775801991,"quota_type":1,"model_ratio":0,"model_price":1,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.1-high","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G2","A1","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-max","description":"Qwen3.7-Max 是阿里云于2026年5月推出的旗舰级通用大语言模型，专为智能体时代打造，旨在提供卓越的逻辑推理、全栈编程及多模态融合能力。该模型支持百万级超长上下文，在复杂长程任务执行与多智能体协作场景中表现出色。目前，用户可通过阿里云百炼平台及相关API接口调用，广泛应用于企业级自动化办公、复杂工程开发及数字化业务落地。","tags":"文本","vendor_id":4,"created_time":1779688225,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.1,"enable_groups":["A1","A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash-image-preview","description":"Google Gemini 2.5 Flash Image预览版，优化了图像理解和生成性能。支持文本驱动图像生成、多图混合、多轮编辑、风格混搭、角色一致性保持和局部编辑功能。具备良好的性价比，适合营销设计师、内容创作者和数字艺术设计人员使用。","icon":"Gemini.Color","tags":"图片","vendor_id":6,"created_time":1766546328,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"nano-banana-pro","description":"Google Gemini 3 Pro Image预览版支持打印级质量输出。具备思考推理能力和搜索Grounding功能，可生成动态内容。支持12+种语言的多语言文字渲染，高级图像混合可融合至14张图像，专注于复杂、多轮、高保真的专业级生成任务。","tags":"图片","vendor_id":6,"created_time":1783837180,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini-image-G1","B1","default"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.0-flash","description":"Google于2024年12月发布的Gemini 2.0系列中速模型。支持多模态输入（文本、代码、图像、音频、视频），最大输入1,048,576 tokens，输出8,192 tokens。具备原生图像生成、原生工具使用、多模态实时API等创新能力。支持模型调优、系统指令、函数调用、结构化输出等功能，平衡速度和性能。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","A1","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash-thinking","description":"Google Gemini 2.5 Flash思考版，启用可见的推理过程展示。通过显示模型的思考步骤，帮助用户理解决策逻辑。专为速度、效率和规模化而优化，支持多模态输入，最大输入1,048,576 tokens，输出65,535 tokens。适合需要透明化推理过程的任务场景。","tags":"文本","vendor_id":6,"created_time":1766666123,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333,"enable_groups":["A2","B1","C2","default","G1","G2","gemini-vertex","A1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro-ci","description":"Google Gemini 2.5 Pro 旗舰推理多模态模型，擅长复杂代码、数学、长文档和视频理解，支持长上下文与工具调用。该名称为服务侧 CI 路由别名，具体底层版本由渠道配置决定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["default","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"grok-4.5-c","description":"xAI Grok 4.5 高能力通用模型，面向复杂推理、编码、实时信息处理和长程智能体任务。当前名称为按次计费别名，实际模型名称为 grok-4.5。","tags":"文本","vendor_id":8,"created_time":1785210665,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.1-thinking-all","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v3.1-search-thinking","description":"DeepSeek V3.1联网搜索思考模式，结合实时联网搜索和深度推理能力。","tags":"联网","vendor_id":3,"created_time":1737782400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwq-plus-latest","description":"QWQ-Plus最新版本，推理优化大模型，提供更强的语言理解和文本生成能力。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.8,"model_price":0,"owner_by":"","completion_ratio":2.5,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o4-mini-all","description":"OpenAI o4-mini全能版本，提供快速成本高效的推理和完整功能支持。在数学、编码和视觉任务中表现出众，适合多场景应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.2-codex","description":"OpenAI GPT-5.2-Codex最新版本（2025年12月发布），目前最先进的智能体编码模型。通过上下文压缩提升长程任务执行能力，在大规模代码变更（重构、迁移）、Windows环境和网络安全能力方面显著增强。支持终端、IDE、云端沙盒、GitHub代码审查和Slack团队协作。","tags":"文本","vendor_id":1,"created_time":1768102992,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G2","A1","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-pro-low","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该条目将推理强度设为低档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":6,"created_time":1771909189,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"veo_3_1-fast-fl","description":"Google Veo 3.1快速版全长模式（Fast Full-Length）。快速生成完整长度视频，平衡速度和质量，适合需要快速产出完整视频内容的场景。","tags":"视频","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"claude-opus-4-20250514-thinking","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":2,"created_time":1766650855,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"o1-2024-12-17","description":"OpenAI o1模型2024年12月17日版本快照，保留了该时间点的模型能力和特性。提供稳定的推理性能，适合需要固定模型版本的生产环境和长期项目。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-2026-02-15-search-thinking","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。该名称固定到 2026 年 2 月 15 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1772075130,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-image-2-high","description":"GPT Image 2 High 是 OpenAI 于 2026 年 4 月发布的一款旗舰级图像生成模型，属于 GPT Image 2 系列的最高画质档位。该模型具备内置推理能力，支持原生 4K 分辨率输出，在复杂空间布局控制及多语言文本渲染方面表现出色，特别适用于专业设计、产品海报、UI 模型及高保真视觉素材的创作。","tags":"图片","vendor_id":1,"created_time":1787026383,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI生图","OpenAI改图"],"billing_labels":["按次计费"]},{"model_name":"gemini-flash-latest-thinking","description":"Google Gemini Flash 系列高速多模态模型，兼顾推理、长上下文、工具调用和成本效率，适合实时及高并发应用。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G1","G2","gemini-vertex","C2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5.1","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。","tags":"文本","vendor_id":1,"created_time":1763611674,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["C2","codex渠道-gpt","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.8-flash","description":"Qwen3.8-Flash（常指 Qwen3.8-Flash-Next）是阿里巴巴 Qwen 团队于 2026 年 8 月发布的系列模型之一，定位为高性价比的多模态混合专家（MoE）模型。该模型拥有 125B 总参数（激活 6B），具备 256K 上下文窗口，主要面向代理式编程、复杂文档处理及工具调用等高并发、长序列应用场景，旨在通过高效架构在保持性能的同时大幅降低推理成本。","tags":"文本","vendor_id":4,"created_time":1788309869,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":3.375,"cache_ratio":0.125,"create_cache_ratio":1.5625,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-pro-2025-10-06","description":"OpenAI GPT-5 系列专业高能力版本，面向复杂推理、科研分析、软件工程和高价值知识工作，强调答案质量与可靠性。该名称固定到 2025 年 10 月 6 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":10.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"seedance-2-5-720p-i2v","description":"支持 720p 分辨率。Seedance 2.5 音视频联合生成模型，仅支持文生视频和图生视频，支持最长 30 秒单次生成。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":1.36,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-6","description":"轻量高速版本，侧重日常对话、高频轻量任务，响应速度更快，成本极低，普通免费用户可在桌面端体验。","tags":"文本","vendor_id":1,"created_time":1790170630,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G2","A1","B1","C2","codex渠道-gpt","default","G1"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI对话（响应式）"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"standard\", p * 10 + c * 50 + cr * 1 + cc * 12.5) : tier(\"long_context\", p * 20 + c * 75 + cr * 1.5 + cc * 25)"},{"model_name":"deepseek-v3.1-search","description":"DeepSeek V3.1联网模式，结合实时联网搜索和深度推理能力。","tags":"联网","vendor_id":3,"created_time":1769580010,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"o1-pro-all","description":"OpenAI o1-pro全能版本，提供最强推理能力和所有高级功能支持。整合视觉输入、函数调用、结构化输出等全部能力，适合最复杂的企业级推理任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"qwen-plus-latest-search-thinking","description":"通义千问Plus最新版联网搜索思考模式，结合联网搜索和深度推理能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":13.333333333333,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-all-c","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","tags":"文本","vendor_id":2,"created_time":1770778732,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-pro-thinking-512","description":"Google Gemini 2.5 Pro 旗舰推理多模态模型，擅长复杂代码、数学、长文档和视频理解，支持长上下文与工具调用。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该别名启用标记为 512 的推理配置，具体推理预算以渠道实现为准。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gemini-3-pro-low-c","description":"Google Gemini 3 Pro低推理强度版，优化响应速度和成本效率。通过thinking_level参数控制推理深度，提供快速而准确的响应。保留Gemini 3 Pro的核心能力，但减少推理时间，适合需要快速响应的常规任务、实时对话和高频查询场景。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.0-flash-lite","description":"Google于2025年2月发布的Gemini 2.0系列最具成本效益和最快的模型。专为大规模文本输出场景优化，具有最低延迟。支持多模态输入（文本、代码、图像、音频、视频），最大输入1,048,576 tokens，输出8,192 tokens。采用单一定价模式，降低长上下文请求成本。支持模型调优、结构化输出、函数调用等功能。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.0375,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-vl-ocr","description":"视觉理解文字识别模型，专门用于从各类图像中提取文本和结构化数据。支持多语言识别、倾斜文本、文本定位、信息提取、表格解析和公式识别。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-search","description":"通义千问旗舰模型联网搜索版本，结合实时联网能力处理复杂任务。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":3.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-o3","description":"OpenAI最强大的推理模型（2025年4月发布），在编码、数学、科学和视觉感知等领域表现突出。在代码竞赛Codeforces得分2727Elo，SWE-bench准确率71.7%（比o1提升20%以上），数学竞赛准确率96.7%。首次实现推理模型综合使用ChatGPT所有工具（网络搜索、Python数据分析、图像分析等），在思维链过程中使用工具增强能力。当前名称为按次计费别名，实际模型名称为 o3。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.25,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"wan2.7-t2v-1080P","description":"阿里通义万相 Wan2.7 文生视频模型，根据文本提示生成具备连贯运动与镜头语言的视频内容，适合短片与商业创作。","tags":"视频","vendor_id":4,"created_time":1775802113,"quota_type":1,"model_ratio":0,"model_price":1,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"cs-claude-opus-4-5-20251101","description":"Claude Opus 4.5 是 Anthropic 于 2025 年 11 月推出的前沿推理模型，专为复杂的软件工程、智能体（Agent）工作流及长周期计算机操作而优化。该模型在代码编写、多步逻辑推理及工具调用方面表现出色，并具备极高的推理效率与上下文处理能力，被广泛应用于自动化研究、深度调试及企业级复杂任务处理。当前名称为按次计费别名，实际模型名称为 claude-opus-4-5-20251101。该名称固定到 2025 年 11 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"o3-mini-all","description":"OpenAI o3-mini全能版本，提供快速推理和完整功能支持。在低成本的同时保持o3-mini的高性能，适合需要经济高效的多场景推理应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.05,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-nothinking","description":"Google Gemini 2.5 Flash标准版，关闭思考过程显示。专为速度、效率和规模化而优化，提供最佳性价比。适用于大规模摘要生成、响应式聊天应用和数据提取等高吞吐量企业任务。支持多模态输入，最大输入1,048,576 tokens，输出65,535 tokens。","tags":"文本","vendor_id":6,"created_time":1766666047,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.33333,"enable_groups":["A1","A2","C2","default","G2","B1","G1","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"o3","description":"OpenAI最强大的推理模型（2025年4月发布），在编码、数学、科学和视觉感知等领域表现突出。在代码竞赛Codeforces得分2727Elo，SWE-bench准确率71.7%（比o1提升20%以上），数学竞赛准确率96.7%。首次实现推理模型综合使用ChatGPT所有工具（网络搜索、Python数据分析、图像分析等），在思维链过程中使用工具增强能力。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v3.2-thinking","description":"DeepSeek V3.2思考模式，支持复杂问题的深度推理分析。","tags":"文本","vendor_id":3,"created_time":1737782400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-preview-search","description":"通义千问旗舰模型预览版联网搜索。","tags":"联网","vendor_id":4,"created_time":1769691283,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"chatgpt-4o-latest","description":"OpenAI ChatGPT 当前 GPT-4o 对话版本的 API 别名，随 ChatGPT 产品更新，侧重自然对话、写作和多模态理解，不保证固定快照行为。","tags":"文本","vendor_id":1,"created_time":1763689608,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-1.5-flash-latest","description":"Google Gemini 1.5 Flash 高速多模态模型，支持长上下文以及文本、图像、音频和视频理解，适合高吞吐内容处理与实时应用。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"Moonshot-Kimi-K2-Instruct-search","description":"Kimi K2联网搜索版本，结合实时联网能力和强大的代码生成、推理能力。","tags":"联网","vendor_id":9,"created_time":1737782400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-4-sonnet","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。当前名称为按次计费别名，实际模型名称为 claude-4-sonnet。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"qwen3-vl-flash-thinking","description":"轻量级多模态视觉-语言模型思考模式，快速响应的同时支持视觉推理。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-plus-2026-05-26","description":"Qwen3.7-Plus-2026-05-26 是 Qwen3.7 Plus 的 2026 年 5 月 26 日快照，全面增强视觉语言理解、编码、工具使用和生产力智能体能力，支持图像、文本及视频输入和约 100 万 token 上下文。","tags":"文本","vendor_id":4,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 256000\n  ? tier(\"standard\", p * 1.6 + c * 6.4 + cr * 0.32)\n  : tier(\"long_context\", p * 4.8 + c * 19.2 + cr * 0.96)"},{"model_name":"cs-claude-sonnet-4-5-thinking-all","description":"Claude Sonnet 4.5 是 Anthropic 于 2025 年 9 月发布的高性能推理模型，主要面向复杂的长链路任务、智能体构建及专业代码编程场景。该模型支持 20 万上下文窗口及多模态视觉输入，曾以卓越的逻辑推理和工具使用能力在编程基准测试中表现突出。随着技术迭代，目前该模型已由更新的版本（如 Sonnet 4.6 及后续系列）接替。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-5-thinking-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-preview","description":"Google Gemini 2.5 Flash 高速多模态模型，在推理能力、响应速度和调用成本之间取得平衡，适合高并发智能体与内容处理。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v4-flash-0731","description":"DeepSeek-V4-Flash-0731 是 2026 年 7 月 31 日发布的高效混合专家模型，总参数 284B、激活参数 13B，原生支持 100 万 token 上下文，兼顾推理质量、低延迟与高吞吐，适合日常对话、内容创作、基础 RAG、编程和智能体任务。","tags":"文本","vendor_id":3,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["C2","default","G1","G2","A1","A2","B1"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"(tier(\"base\", p * 1.5 + c * 4.5 + cr * 0.05)) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 9 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 12 ? 2 : 1) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 14 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 18 ? 2 : 1)"},{"model_name":"gemini-3.1-flash-lite-preview","description":"Google Gemini 3.1 Flash-Lite 轻量高效多模态模型，针对低延迟、低成本和大规模生产负载优化，适合分类、抽取与实时交互。","tags":"文本","vendor_id":6,"created_time":1772630017,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","A2","B1","G1","G2","gemini-vertex","C2","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-3.5-turbo-1106","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。该名称固定到 2023 年 11 月 6 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v4-pro-0813","description":"DeepSeek-V4-Pro-0813 是 DeepSeek 于 2026 年 8 月发布的混合专家（MoE）模型，作为 DeepSeek-V4-Pro 的正式版本，该模型拥有 1.6 万亿参数（490 亿激活）。它在增强代理能力与生产环境性能的基础上，集成了 DSpark 推理加速模块，并支持 100 万 token 的上下文窗口，主要面向复杂的代码开发、工具调用及长周期代理任务。","tags":"文本","vendor_id":3,"created_time":1787112942,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["C2","default","G1","G2","A1","A2","B1"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"(tier(\"base\", p * 4.5 + c * 13.5 + cr * 0.15)) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 9 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 12 ? 2 : 1) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 14 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 18 ? 2 : 1)"},{"model_name":"cs-grok-4-1","description":"xAI Grok 4.1 高能力通用模型，强化推理、代码、工具调用和智能体任务，在复杂工作流中兼顾准确性与执行效率。当前名称为按次计费别名，实际模型名称为 grok-4-1。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.07,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"qwen-long","description":"上下文窗口最长的通义千问模型，专为长文本处理优化，推理能力较强。适合长文本对话、分析、信息抽取、总结摘要等任务。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G2","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-o3-all","description":"OpenAI o3全能版本，整合最强推理模型的所有能力和工具集成。提供完整的网络搜索、Python数据分析、图像分析等功能，适合需要全面o3能力的复杂应用。当前名称为按次计费别名，实际模型名称为 o3-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.25,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-pro-preview-thinking","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","default","A1","A2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1","description":"OpenAI GPT-4.1 旗舰非推理模型，针对编码、指令遵循和长上下文理解优化，适合软件开发、文档分析和工具调用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-chat-2025-10-03","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该名称固定到 2025 年 10 月 3 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-6-astra","description":"GPT-6 Astra 是 OpenAI 于 2026 年 9 月推出的前沿大语言模型，定位为能够像计算机操作员一样自主完成复杂任务的智能系统。该模型在计算机使用、软件工程、科学研究及网络安全等领域表现出色，具备强大的多步工作流处理与屏幕交互能力。其支持 105 万 token 的上下文窗口，主要面向开发者与企业用户，旨在通过增强的意图理解与工具调用能力，实现更高水平的自动化作业。","tags":"文本","vendor_id":1,"created_time":1788578839,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["A1","A2","B1","default","C2","codex渠道-gpt","G1","G2"],"supported_endpoint_types":["openai","OpenAI对话","OpenAI对话（响应式）"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"standard\", p * 10 + c * 50 + cr * 1 + cc * 12.5) : tier(\"long_context\", p * 20 + c * 75 + cr * 1.5 + cc * 25)"},{"model_name":"claude-opus-4-8-thinking-c","description":"Claude-Opus-4-8-Thinking 是指 Anthropic 于 2026年5月28日 发布的旗舰模型 Claude Opus 4.8 在开启思考（Thinking/Reasoning）模式下的运行形态，它是一款混合推理模型，通过自适应思考（Adaptive Thinking）与 effort 参数（low/medium/high/xhigh/max）动态控制推理深度，专为复杂推理、长周期自主智能体编码及高阶专业知识工作设计，默认支持 1M Token 上下文窗口与最高 128K 输出。","tags":"文本","vendor_id":2,"created_time":1780542015,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"cs-gpt-4.1-nano","description":"OpenAI GPT-4.1 系列最轻量模型，提供低延迟和低成本的长上下文处理，适合分类、抽取、自动补全及高吞吐任务。当前名称为按次计费别名，实际模型名称为 gpt-4.1-nano。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.0012,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"net-o1-mini-all","description":"OpenAI o1-mini全能版本，提供o1-mini的快速推理能力和全面功能支持。适合需要经济高效的STEM推理和通用编程任务。当前名称为服务侧路由别名，底层模型为 o1-mini-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-thinking-*","description":"Google Gemini 2.5 Flash 高速多模态模型，在推理能力、响应速度和调用成本之间取得平衡，适合高并发智能体与内容处理。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。名称保留渠道提供的字面量 * 后缀，用于其自身的路由约定。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333,"enable_groups":["G1","G2","gemini-vertex","C2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-lite","description":"Gemini-3.1-flash-lite​ 是 Google DeepMind 于 2026年3月​ 发布的 Gemini 3 系列​ 中速度最快、成本最低的轻量化模型，专为高并发、低延迟、成本敏感的大规模生产级工作负载设计。它并非简单的“阉割版”，而是基于 Gemini 3 Pro 架构​ 针对吞吐量与延迟深度优化的高性价比推理模型","tags":"文本","vendor_id":6,"created_time":1781658542,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["gemini-vertex","A1","A2","C2","default","G1","G2","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"glm-5.1","description":"智谱 GLM-5.1 新一代高能力模型，面向复杂推理、编程和长程智能体任务优化，强化工具调用与生产级任务执行。","tags":"文本","vendor_id":7,"created_time":1776302616,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-3.5-turbo-16k-0613","description":"OpenAI 面向高吞吐对话场景推出的 GPT-3.5 系列模型，擅长日常问答、文本生成、摘要、分类和基础代码辅助，具有响应快、成本低的特点。该名称固定到 2023 年 6 月 13 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G2","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-6-thinking","description":"Claude Opus 4.6 思考版 是 Anthropic 于 2026 年 2 月 6 日发布的最新旗舰 AI 模型‌，支持 100 万 token 上下文窗口和智能体协作功能，在编程、金融分析及多任务处理能力上实现突破性升级。‌‌\n","tags":"文本","vendor_id":2,"created_time":1770556700,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-chat","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.2-2025-12-11","description":"OpenAI GPT-5.2 专业通用模型，强化复杂推理、编码、工具使用和长时间任务执行，适合生产级知识工作与智能体应用。该名称固定到 2025 年 12 月 11 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1767605119,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G1","G2","A1","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-sol","description":"GPT-5.6 Sol 发布于 2026 年 6 月 26 日，是 GPT-5.6 系列中的旗舰模型，也是该系列能力最强的版本。它面向复杂推理、软件工程、科研分析、专业知识工作、计算机使用和网络安全等高难度任务，适合需要深度思考、长链路规划、工具协作和高质量输出的场景。Sol 支持更高强度的推理设置，适合处理复杂代码任务、长期研究任务、专业分析和高价值生产工作流。","tags":"文本","vendor_id":1,"created_time":1783653912,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","C2","codex渠道-gpt","default","G1","G2","A1","A2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5 + cc * 6.25) : tier(\"long_context\", p * 10 + c * 45 + cr * 1 + cc * 12.5)"},{"model_name":"qwen3-vl-flash-2025-10-15-search-thinking","description":"轻量级多模态视觉-语言模型2025年10月15日版本联网搜索思考模式。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-20250514","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-20250514。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-pro-preview-03-25","description":"Google Gemini 2.5 Pro 旗舰推理多模态模型，擅长复杂代码、数学、长文档和视频理解，支持长上下文与工具调用。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-image-preview-2k","description":"Nano Banana 2（官方标识为 gemini-3.1-flash-image-preview）是 Google DeepMind 发布的新一代 AI 图像生成与编辑模型。该模型基于 Gemini 3.1 Flash Image 架构打造，在保持 Flash 级别极速生成的同时，大幅缩小了与 Pro 版本在画质上的差距，实现了速度、质量与成本的高效平衡。","tags":"图片","vendor_id":6,"created_time":1772158054,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini-image-G1","B1","default"],"supported_endpoint_types":["openai","image-edits","image-generation"],"billing_labels":["按次计费"]},{"model_name":"qwen3-vl-flash-search","description":"轻量级多模态视觉-语言模型联网搜索版本，快速视觉理解结合实时联网。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-gemini-2.5-flash-all","description":"Google Gemini 2.5 Flash全功能版，集成所有可用工具和能力。专为速度、效率和规模化而优化，提供最佳性价比。支持多模态输入（文本、代码、图像、音频、视频），具备思考能力、Google搜索整合、函数调用、结构化输出、上下文缓存等完整高级功能集。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.016,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-0613","description":"OpenAI GPT-4 系列高能力通用模型，擅长复杂指令遵循、逻辑推理、代码生成和专业文本处理，是早期 GPT-4 代旗舰模型。该名称固定到 2023 年 6 月 13 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.5-2026-04-24","description":"OpenAI GPT-5.5 前沿通用模型，面向复杂编码、研究分析和真实业务任务，强化长链路推理、工具协作与高质量专业输出。该名称固定到 2026 年 4 月 24 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"happyhorse-1.0-i2v-1080P","description":"首帧图生视频","tags":"视频","vendor_id":4,"created_time":1780043356,"quota_type":1,"model_ratio":0,"model_price":1.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["default","B1"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.4-pro","description":"OpenAI GPT-5.4 系列高能力专业模型，面向高难度推理、软件工程、研究和长程智能体任务，强调准确性与任务完成质量。","tags":"文本","vendor_id":1,"created_time":1773729425,"quota_type":0,"model_ratio":15,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-3-search","description":"xAI Grok 3 通用大语言模型，擅长推理、代码、实时信息分析和工具使用，适合对话、研究与复杂问题求解。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-image-2.0-pro","description":"阿里通义千问 Qwen-Image 2.0 Pro 专业图像生成与编辑模型，强化真实感、复杂构图、文字渲染和多图参考一致性，适合商业设计。","tags":"图片","vendor_id":4,"created_time":1773899359,"quota_type":1,"model_ratio":0,"model_price":0.5,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","image-generation"],"billing_labels":["按次计费"]},{"model_name":"cs-gemini-2.5-pro-all","description":"Google Gemini 2.5 Pro全功能版，集成所有可用工具和能力。旗舰级AI模型，具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。支持原生音频输出、高级安全防护、Project Mariner计算机使用功能、\"Deep Think\"增强推理模式等完整高级功能集。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按次计费"]},{"model_name":"gemini-1.5-pro-latest","description":"Google Gemini 1.5 Pro 高能力多模态模型，擅长长文档、代码、图像、音频和视频理解，适合复杂分析与长上下文任务。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","B1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-flash-preview-05-20","description":"Google Gemini 2.5 Flash预览版（2025年5月20日版本），提供早期访问体验。专为速度、效率和规模化而优化，提供最佳性价比。支持多模态输入（文本、代码、图像、音频、视频），最大输入1,048,576 tokens，输出65,535 tokens。具备思考能力、Google搜索整合、函数调用等高级功能。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":8.333,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"happyhorse-1.0-t2v-720P","description":"文生视频","tags":"视频","vendor_id":4,"created_time":1780043200,"quota_type":1,"model_ratio":0,"model_price":0.9,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"claude-sonnet-5-5","description":"2026 年 9 月 28 日正式推出的 Claude 5.5 系列中端主力大模型，API 标识为`claude-sonnet-5-5`，作为旗舰 Claude Opus 5.5 的配套主力模型，主打**能力、速度与成本三者平衡**，面向规模化业务落地、Agent 智能体开发、代码工程、文档处理等高频生产场景。","tags":"文本","vendor_id":2,"created_time":1790656603,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["C2","Claude Max","Claude-Aws","default","G1","G2"],"supported_endpoint_types":["anthropic","openai","OpenAI对话","Anthropic"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-all","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.4-nano-2026-03-17","description":"OpenAI GPT-5.4 系列最轻量模型，面向分类、抽取、简单生成和大规模低延迟处理，强调速度与成本效率。该名称固定到 2026 年 3 月 17 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":6.25,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-5-5","description":"Anthropic 于 2026 年 9 月 22 日发布的旗舰级大模型，为 Claude 5.5 系列首款产品，主打**长周期智能体编码、专业知识工作**，在绝大多数任务上性能对标 Anthropic 最强模型 Claude Fable 5.1，同时相比上一代 Opus 5，综合运行成本降低 40%，生成速度提升 30% 以上。","tags":"文本","vendor_id":2,"created_time":1790170682,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["G1","G2","default","A1","A2","B1","C2","Claude Max","Claude-Aws"],"supported_endpoint_types":["anthropic","openai","OpenAI对话","Anthropic"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 4 + c * 20 + cr * 0.2 + cc * 5 + cc1h * 8)"},{"model_name":"gpt-5-mini-2025-08-07","description":"OpenAI GPT-5 系列轻量模型，兼顾推理能力、速度和成本，适合高并发对话、代码辅助、摘要与结构化数据处理。该名称固定到 2025 年 8 月 7 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o4-mini-high-all","description":"OpenAI o4-mini-high全能版本，提供o4-mini系列的高性能配置和全面功能。在保持成本效益的同时提供更高的推理质量，适合需要平衡性能和成本的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-32k","description":"OpenAI GPT-4 系列高能力通用模型，擅长复杂指令遵循、逻辑推理、代码生成和专业文本处理，是早期 GPT-4 代旗舰模型。","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":30,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G1","G2","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-pro-preview","description":"Google于2025年11月发布的最先进推理模型预览版。具备前所未有的推理深度和细微差别理解能力，提供智能、简洁、直接的响应。在高级推理、复杂指令遵循、工具使用和代理用例方面显著超越Gemini 2.5 Pro。支持1百万token上下文窗口（约1500页文本或30000行代码），世界领先的多模态理解能力。","tags":"文本","vendor_id":6,"created_time":1763611674,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["G2","gemini-vertex","A1","A2","B1","C2","default","G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.1-codex-max","description":"OpenAI GPT-5.1-Codex最高性能版本，提供最强的代码理解和生成能力。适合大规模代码变更、架构重构、复杂系统设计等高难度编程任务，在代码质量和智能程度上达到最高水平。","tags":"文本","vendor_id":1,"created_time":1766546328,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max-search-thinking","description":"通义千问旗舰模型联网搜索思考模式，结合联网搜索和深度推理能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":3.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.7-plus","description":"Qwen3.7-Plus 是由阿里巴巴通义千问团队推出的一款高性能大语言模型，旨在平衡推理能力与算力成本。该模型在多语言理解、复杂逻辑推理及长文本处理方面表现出色，并针对智能体（Agent）场景进行了专项优化。目前该模型通过阿里云平台提供服务，被广泛应用于企业级生产力应用、自动化办公及多样化的开发者工具构建。","tags":"文本","vendor_id":4,"created_time":1784126508,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 256000\n  ? tier(\"standard\", p * 1.6 + c * 6.4 + cr * 0.32)\n  : tier(\"long_context\", p * 4.8 + c * 19.2 + cr * 0.96)"},{"model_name":"gemini-2.0-flash-exp","description":"Google Gemini 2.0 Flash 高速多模态模型，支持文本、图像、音频和视频理解，并具备长上下文、工具调用与结构化输出能力。","tags":"文本","vendor_id":6,"created_time":1766658068,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-image-2.0-pro-2026-03-03","description":"阿里通义千问 Qwen-Image 2.0 Pro 专业图像生成与编辑模型，强化真实感、复杂构图、文字渲染和多图参考一致性，适合商业设计。该名称固定到 2026 年 3 月 3 日版本，适合需要版本一致性的应用。","tags":"图片","vendor_id":4,"created_time":1773899345,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","image-generation"],"billing_labels":["按次计费"]},{"model_name":"o3-mini-2025-01-31-medium","description":"OpenAI o3-mini 轻量推理模型，擅长数学、科学、代码和多步骤问题求解，并可通过推理强度配置平衡速度与答案质量。该条目将推理强度设为中档，以在速度、成本和回答质量之间提供固定配置。该名称固定到 2025 年 1 月 31 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-mini","description":"OpenAI GPT-4o 系列轻量多模态模型，支持文本与图像理解，具有速度快、成本低的特点，适合高并发对话和结构化处理。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-codex","description":"OpenAI GPT-5系列代码生成模型，统一的智能体编程助手。支持终端和IDE中的配对编程（VSCode、Cursor、Windsurf），云端沙盒环境执行任务，GitHub自动代码审查，Slack团队协作集成。CLI和IDE扩展推荐作为默认编码模型。","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G2","A1","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-2025-04-28","description":"通义千问Turbo 2025年4月28日版本，特定日期快照版本。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-plus-2025-12-19-search-thinking","description":"多模态视觉-语言模型2025年12月19日版本联网搜索思考模式。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-8-thinking","description":"Anthropic Claude Opus 4.8 旗舰模型，面向复杂推理、长周期自主智能体编程与企业级知识工作，强化代码库理解和多阶段任务执行。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1785210665,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.0-pro-exp-02-05","description":"Google Gemini 2.0 Pro 实验性高能力多模态模型，面向复杂推理、代码和长上下文分析，用于验证该系列的旗舰能力。该名称为实验版本，主要用于提前验证新能力。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"suno_concat","description":"Suno AI音乐拼接与扩展功能，将多个音乐片段无缝连接或延长现有音乐。支持从2分钟基础版本扩展到4分钟完整版本（V3+），保持音乐风格一致性，适合创作完整长曲目。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-pro-preview","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。","tags":"文本","vendor_id":6,"created_time":1771744637,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["A1","A2","B1","default","G1","C2","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-20250514","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1762411850,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["B1","default","A1","A2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-terra-openai-compact","description":"GPT-5.6 Terra 的 OpenAI Responses Compact 路由别名，平衡质量、延迟与成本，用于压缩长对话状态，实际底层模型由渠道配置决定。","tags":"文本","vendor_id":1,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["A1","B1","C2","codex渠道-gpt","default","G1","G2"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 2.5 + c * 15 + cr * 0.25 + cc * 0.125) : tier(\"long_context\", p * 5 + c * 22.5 + cr * 0.5 + cc * 0.25)"},{"model_name":"o4-mini-2025-04-16","description":"OpenAI o4-mini模型2025年4月16日版本，提供稳定的快速推理能力。在数学和编码任务中表现出众，适合需要快速且经济的推理解决方案。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"deepseek-v3.1-thinking","description":"DeepSeek V3.1思考模式，支持混合模式，可通过改变聊天模板在思考和非思考模式间切换。","tags":"文本","vendor_id":3,"created_time":1737782400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"claude-haiku-4-5-20251001-thinking","description":"Anthropic Claude Haiku 4.5 高速高性价比模型，强化编码、智能体和工具调用能力，适合实时交互与大规模生产工作负载。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 10 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["default","B1"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-image-4k","description":"Gemini 3.1 Flash Image（代号 Nano Banana 2）是 Google 于 2026 年 6 月正式发布的图像生成与编辑模型。该模型专注于高效率与低延迟，支持通过自然语言进行对话式图像编辑、多角色一致性保持及实时参考引用，主要面向高频次、大规模的图像生成与创意工作流场景。","tags":"图片","vendor_id":6,"created_time":1785055034,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["openai","image-edits","image-generation"],"billing_labels":["按次计费"]},{"model_name":"gpt-5-2025-08-07","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该名称固定到 2025 年 8 月 7 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o3-2025-04-16","description":"OpenAI o3模型2025年4月16日官方发布版本，标志着o3系列正式上线。提供完整的o3推理能力和工具集成，适合需要最强推理性能的生产环境。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2","A1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-max","description":"通义千问旗舰文本模型，系列效果最好，适合复杂、多步骤任务。支持256K超大上下文，具备函数调用、结构化输出、推理模式等核心功能。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":3.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-gemini-3-pro-all","description":"Google Gemini 3 Pro全功能版，集成所有可用工具和能力。最先进的推理模型，具备前所未有的推理深度、世界领先的多模态理解、1百万token上下文窗口。集成Canvas创意工作空间、Storybook故事创作、深度搜索、思考可视化等完整高级功能集。适合需要全方位AI能力的复杂场景。","icon":"Gemini.Color","tags":"文本","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按次计费"]},{"model_name":"o1-preview","description":"OpenAI o1系列早期预览版本，设计用于利用广泛通用知识解决复杂问题。支持更广泛的STEM以外领域的世界知识，最大输出限制32,768 tokens。适合法律分析、科研推理、金融决策等需要高精度逻辑和博士级推理能力的场景。成本15美元/百万输入tokens。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["default","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"grok-4.1","description":"xAI Grok 4.1 高能力通用模型，强化推理、代码、工具调用和智能体任务，在复杂工作流中兼顾准确性与执行效率。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen-long-2025-01-25","description":"通义千问Long 2025年1月25日版本，超长上下文处理能力。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3-vl-plus-2025-12-19","description":"多模态视觉-语言模型2025年12月19日版本。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-image-preview","description":"‌Gemini 3.1 Flash Image Preview‌（又称“Nano Banana 2”）是 Google 推出的轻量级多模态图像生成模型，专为‌高速、高吞吐、低成本‌的文生图（Text-to-Image）任务设计，在保持 Pro 级画质的同时显著提升效率 ‌。","tags":"图片","vendor_id":6,"created_time":1772158031,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v3.2-search","description":"DeepSeek V3.2联网搜索版本，结合实时联网能力处理问题。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"文本","vendor_id":3,"created_time":1768658421,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["A2","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-turbo-thinking","description":"通义千问Turbo思考模式，快速响应的同时具备推理分析能力。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"qwen3.5-plus-2026-02-15-thinking","description":"阿里通义千问 Qwen3.5-Plus 均衡型通用模型，强化推理、代码、长上下文和智能体能力，适合企业应用与复杂多步骤任务。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2026 年 2 月 15 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1772075142,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-medium","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该条目将推理强度设为中档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G2","A1","B1","C2","default","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"kimi-k2.5-thinking","description":"Kimi-K2.5 （思考版）是月之暗面（Moonshot AI）于 2026 年 1 月 27 日正式发布并开源的新一代全能型开源大模型，被定义为 Kimi 系列迄今最智能、综合能力最强的模型。","tags":"文本","vendor_id":9,"created_time":1770344764,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5.25,"enable_groups":["B1","C2","default","G1","G2","A2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.4-openai-compact","description":"OpenAI GPT-5.4 的上下文压缩路由，用于通过 Responses Compact 接口压缩长对话状态，降低后续请求的上下文负担。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 2.5 + c * 15 + cr * 0.25) : tier(\"long_context\", p * 5 + c * 22.5 + cr * 0.5)"},{"model_name":"gemini-3.1-pro-preview-low","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该条目将推理强度设为低档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":6,"created_time":1771912487,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-flash-lite-latest","description":"Google Gemini Flash-Lite 轻量多模态模型，强调低延迟、低成本和大规模吞吐，适合分类、抽取与快速对话。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"MiniMax-M2.5","description":"MiniMax-M2.5是MiniMax推出的最新一代AI模型，旨在解决现实世界中的复杂任务，编程与智能体能力达到或超越 Opus 4.6 水平。","tags":"文本","vendor_id":34,"created_time":1770952695,"quota_type":0,"model_ratio":1.05,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["G1","G2","A2","B1","C2","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-plus-search","description":"通义千问Plus联网搜索版本，结合实时联网能力处理问题。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":2.4,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-flash-image","description":"Gemini 3.1 Flash Image（代号 Nano Banana 2）是 Google 于 2026 年 6 月正式发布的图像生成与编辑模型。该模型专注于高效率与低延迟，支持通过自然语言进行对话式图像编辑、多角色一致性保持及实时参考引用，主要面向高频次、大规模的图像生成与创意工作流场景。","tags":"图片","vendor_id":6,"created_time":1785054963,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"net-gpt-4o-mini","description":"OpenAI GPT-4o 系列轻量多模态模型，支持文本与图像理解，具有速度快、成本低的特点，适合高并发对话和结构化处理。当前名称为服务侧路由别名，底层模型为 gpt-4o-mini。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.006,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"o3-pro","description":"OpenAI史上最强AI模型（2025年6月发布），在数学基准AIME 2024中超越Google Gemini 2.5 Pro，在博士级科学测试GPQA Diamond中击败Anthropic Claude 4 Opus。可访问所有工具：网络搜索、文件分析、Python使用、视觉推理等。o3整体降价80%，提供卓越性价比。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai-response"],"billing_labels":["按量计费"]},{"model_name":"veo3.1-fast","description":"Google Veo 3.1快速版本，在Veo 3.1基础上优化生成速度。保持改进的提示词追随性和角色一致性，同时提供更快的响应时间，适合需要快速迭代的创意项目。","tags":"视频","vendor_id":6,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"glm-5.3","description":"GLM-5.3 是智谱 AI（Z.ai）于 2026 年 8 月发布的旗舰模型，基于 GLM-5.2 的基座并通过大规模后训练实现性能跃升。该模型在复杂编程任务中表现卓越，编程能力较前代提升 50%，并在漏洞挖掘等网络安全领域展现出显著的涌现能力。模型支持 100 万 token 的长上下文窗口，主要面向复杂软件工程与长程智能体任务，目前已通过相关编码计划提供服务。","tags":"文本","vendor_id":7,"created_time":1787112800,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"enable_groups":["G1","G2","A1","A2","B1","C2","default"],"supported_endpoint_types":["openai","OpenAI对话"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-6-thinking-c","description":"Anthropic Claude Sonnet 4.6 高性能均衡模型，面向复杂编码、长程智能体、工具使用和专业知识工作，在速度与能力之间取得平衡。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-6-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1771926450,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"cs-grok-3","description":"xAI Grok 3 通用大语言模型，擅长推理、代码、实时信息分析和工具使用，适合对话、研究与复杂问题求解。当前名称为按次计费别名，实际模型名称为 grok-3。","icon":"XAI.Color","tags":"文本","vendor_id":8,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.005,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-pro-c","description":"Google Gemini 2.5 Pro 旗舰推理多模态模型，擅长复杂代码、数学、长文档和视频理解，支持长上下文与工具调用。当前名称为按次计费别名，实际模型名称为 gemini-2.5-pro。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"claude-fable-5-1","description":"Anthropic 于 2026 年 9 月推出的 Claude Fable 5.1 是其最新的前沿模型，专为处理复杂的长周期编程、科学研究及企业级智能代理任务而设计。该模型具备 100 万 token 上下文窗口及自适应思维能力，在深度推理任务中表现卓越，并能通过内置的自我验证机制显著降低错误输出，是目前 Anthropic 针对高难度知识工作打造的最强模型之一。","tags":"文本","vendor_id":2,"created_time":1788320356,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["Claude Max","Claude-Aws","G1","A1","default","G2","A2","B1","C2"],"supported_endpoint_types":["anthropic","openai","OpenAI对话","Anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.0-flash-search","description":"Google Gemini 2.0 Flash 高速多模态模型，支持长上下文、工具调用、结构化输出以及文本、图像、音频和视频理解。该版本集成联网检索，可结合实时信息完成问答、研究和资料汇总。","tags":"联网","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"gpt-5.1-2025-11-13","description":"OpenAI GPT-5.1 通用模型，在指令遵循、编码、推理和对话自然度方面进一步优化，适合复杂业务流程、开发辅助和智能体任务。该名称固定到 2025 年 11 月 13 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1767605108,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G1","G2","A1","B1","C2","default"],"supported_endpoint_types":["gemini","openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"MiniMax-M3","description":"MiniMax-M3是MiniMax于2026年6月1日发布的开放权重原生多模态旗舰大模型，采用约428B总参数、23B激活参数的MoE架构与自研MiniMax Sparse Attention（MSA）稀疏注意力机制，API最高支持1M tokens超长上下文窗口（保障至少512K可用）且在1M规模下单token计算量仅为上一代约1/20、预填充与解码速度分别提升9倍与15倍以上，从训练起始阶段即进行文本、图像、视频混合多模态训练以实现深层语义对齐并具备桌面操作能力，在SWE-Bench Pro（59.0%）、BrowseComp（83.5分超越Opus 4.7）、Terminal-Bench 2.1（66.0%）等编码与智能体评测中达到国际前沿水平，已于2026年6月12日正式开源并适配华为昇腾等算力平台","tags":"文本","vendor_id":34,"created_time":1784110431,"quota_type":0,"model_ratio":2.1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.2,"enable_groups":["A2","B1","default"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen3.6-plus-2026-04-02-thinking","description":"阿里通义千问 Qwen3.6-Plus 高性能通用模型，进一步增强复杂推理、编码、工具调用和长程智能体能力，兼顾质量与调用效率。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2026 年 4 月 2 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":4,"created_time":1775140286,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"seedance-2-5-480p-i2v","description":"支持 480p 分辨率。Seedance 2.5 音视频联合生成模型，仅支持文生视频和图生视频，支持最长 30 秒单次生成。","tags":"视频","vendor_id":5,"created_time":1786443762,"quota_type":1,"model_ratio":0,"model_price":0.61,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"suno_act_mp4","description":"Suno AI音乐视频生成功能，将生成的音乐转换为MP4格式的音乐视频。支持根据音乐风格自动生成个性化封面设计，适合社交媒体分享、音乐推广和视觉化展示。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","Suno任务详情","Suno提交","Suno任务查询"],"billing_labels":["按次计费"]},{"model_name":"suno_lyrics","description":"Suno AI歌词生成功能（ReMi AI歌词助手），由AI驱动自动创作歌词。用户只需描述音乐风格和主题，即可一键生成大胆、创意十足的歌词。支持多语言（包括中文），适合快速创作和灵感激发。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-pro","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。","tags":"文本","vendor_id":6,"created_time":1771912440,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G2","gemini-vertex","A1","B1","default","A2","C2","G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"seedance-2-0-1080p-i2v","description":"支持 1080p 分辨率。Seedance 2.0 标准视频生成模型，仅支持文生视频和图生视频，适合多镜头叙事、电影感镜头与音画同步创作。","tags":"视频","vendor_id":5,"created_time":1786853744,"quota_type":1,"model_ratio":0,"model_price":2.25,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI视频"],"billing_labels":["按次计费"]},{"model_name":"suno_uploads","description":"Suno AI音频上传与覆盖功能（Covers功能V4），上传现有音频并输入提示词，生成全新演绎版本。支持音乐翻唱、重新编曲、风格转换。可将旧版本音乐通过Remaster功能升级为V4高音质，修复噪音和失真问题。适合音乐改编、版本升级、创意重混。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","Suno任务查询","Suno任务详情","Suno提交"],"billing_labels":["按次计费"]},{"model_name":"gemini-flash-lite-latest-nothinking","description":"Google Gemini Flash-Lite 轻量多模态模型，强调低延迟、低成本和大规模吞吐，适合分类、抽取与快速对话。该版本不启用扩展思考，优先提供更低延迟和更高吞吐量。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"claude-sonnet-4-5-thinking-all","description":"Claude Sonnet 4.5 是 Anthropic 于 2025 年 9 月发布的高性能推理模型，主要面向复杂的长链路任务、智能体构建及专业代码编程场景。该模型支持 20 万上下文窗口及多模态视觉输入，曾以卓越的逻辑推理和工具使用能力在编程基准测试中表现突出。随着技术迭代，目前该模型已由更新的版本（如 Sonnet 4.6 及后续系列）接替。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","tags":"文本","vendor_id":2,"created_time":1766657929,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-sonnet-4-all","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["G1","G2","B1","C2","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3.1-flash-preview","description":"Google Gemini 3.1 Flash 高速多模态模型，兼顾推理、编码、长上下文与成本效率，适合实时智能体和高并发应用。","tags":"文本","vendor_id":6,"created_time":1772591320,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","G1","G2","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o","description":"OpenAI GPT-4o（Omni）多模态旗舰模型，支持文本、图像和音频相关任务，兼顾响应速度与综合能力，适合对话、视觉理解和工具调用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411851,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["G1","G2","A1","B1","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3-flash-c","description":"Google Gemini 3 Flash 高速多模态模型，在速度、成本与推理能力之间取得平衡，适合实时应用、代码辅助和大规模内容处理。当前名称为按次计费别名，实际模型名称为 gemini-3-flash。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["无"]},{"model_name":"claude-opus-4-5","description":"Claude Opus 4.5 是 Anthropic 于 2025 年 11 月推出的前沿推理模型，专为复杂的软件工程、智能体（Agent）工作流及长周期计算机操作而优化。该模型在代码编写、多步逻辑推理及工具调用方面表现出色，并具备极高的推理效率与上下文处理能力，被广泛应用于自动化研究、深度调试及企业级复杂任务处理。","tags":"文本","vendor_id":2,"created_time":1785053789,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["A1","A2","B1","C2","Claude Max","default","G1","G2"],"supported_endpoint_types":["anthropic","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-5-20251101","description":"Claude Opus 4.5 是 Anthropic 于 2025 年 11 月推出的前沿推理模型，专为复杂的软件工程、智能体（Agent）工作流及长周期计算机操作而优化。该模型在代码编写、多步逻辑推理及工具调用方面表现出色，并具备极高的推理效率与上下文处理能力，被广泛应用于自动化研究、深度调试及企业级复杂任务处理。该名称固定到 2025 年 11 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766546328,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["Claude-Aws","A2","C2","default","G1","G2","A1","B1","Claude Max"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.5-flash","description":"Gemini 3.5 Flash 是 Google 于 2026 年 5 月发布的 Gemini 3 系列多模态模型，主要面向智能体（Agent）开发与复杂编码任务。该模型支持 100 万 token 上下文窗口，具备出色的推理与长序列任务处理能力，并在保持 Flash 系列高运行速度的同时，提供了媲美旗舰模型的性能表现。","tags":"文本","vendor_id":6,"created_time":1779336740,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["A1","B1","G1","G2","A2","C2","default","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.3-codex","description":"OpenAI GPT-5.3 系列智能体编码模型，面向复杂软件工程、跨文件修改、长时间任务、终端操作和代码审查优化，适合自主开发工作流。","tags":"文本","vendor_id":1,"created_time":1770563493,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5.6-openai-compact","description":"GPT-5.6 的 OpenAI Responses Compact 路由别名，用于压缩长对话状态并降低后续请求的上下文负担，实际能力由渠道配置决定。","tags":"文本","vendor_id":1,"created_time":1786443762,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["C2","codex渠道-gpt","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","openai-response-compact"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5 + cc * 6.25) : tier(\"long_context\", p * 10 + c * 45 + cr * 1 + cc * 12.5)"},{"model_name":"qwen3-vl-plus-search","description":"多模态视觉-语言模型联网搜索版本，强大的视觉理解结合实时联网能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-1-20250805-thinking","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。当前名称为按次计费别名，实际模型名称为 claude-opus-4-1-20250805-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 8 月 5 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"happyhorse-1.0-t2v-1080P","description":"文生视频","tags":"视频","vendor_id":4,"created_time":1780043226,"quota_type":1,"model_ratio":0,"model_price":1.6,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"deepseek-v4-flash","description":"DeepSeek-V4-Flash 是 DeepSeek 推出的高效能混合专家模型（MoE），拥有 284B 总参数（13B 激活参数）并支持 100 万 token 的上下文窗口。该模型旨在平衡推理能力与响应速度，通过混合注意力机制和结构优化，在编程及智能体任务中表现出色，是兼具高性价比与快速推理的实用选择。","tags":"文本","vendor_id":3,"created_time":1778230091,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["G2","A1","A2","B1","C2","default","G1"],"supported_endpoint_types":["openai"],"billing_labels":["无"],"billing_mode":"tiered_expr","billing_expr":"(tier(\"base\", p * 1.5 + c * 4.5 + cr * 0.05)) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 9 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 12 ? 2 : 1) * (weekday(\"Asia/Shanghai\") \u003e= 1 \u0026\u0026 weekday(\"Asia/Shanghai\") \u003c 6 \u0026\u0026 hour(\"Asia/Shanghai\") \u003e= 14 \u0026\u0026 hour(\"Asia/Shanghai\") \u003c 18 ? 2 : 1)"},{"model_name":"qwen-flash-search-thinking","description":"通义千问Flash联网搜索思考模式，快速响应结合联网搜索和推理能力。","tags":"联网","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["C2","default","G1","G2","A2","B1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-thinking-all","description":"Anthropic Claude Opus 4 旗舰模型，擅长高级编码、复杂推理、研究分析和长时间智能体任务，适合高要求专业场景。当前名称为按次计费别名，实际模型名称为 claude-opus-4-thinking-all。该条目为全功能路由版本，核心模型能力与对应基础模型一致。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"cs-claude-sonnet-4-20250514","description":"Anthropic Claude Sonnet 4 高性能均衡模型，擅长编码、复杂推理、智能体和工具调用，适合软件工程与企业级知识工作。当前名称为按次计费别名，实际模型名称为 claude-sonnet-4-20250514。该名称固定到 2025 年 5 月 14 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-preview-04-17","description":"Google Gemini 2.5 Flash 高速多模态模型，在推理能力、响应速度和调用成本之间取得平衡，适合高并发智能体与内容处理。该名称为预览版本，能力和行为可能随服务商更新而调整。","tags":"文本","vendor_id":6,"created_time":1785210665,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"o3-mini-2025-01-31-low","description":"OpenAI o3-mini 轻量推理模型，擅长数学、科学、代码和多步骤问题求解，并可通过推理强度配置平衡速度与答案质量。该条目将推理强度设为低档，以在速度、成本和回答质量之间提供固定配置。该名称固定到 2025 年 1 月 31 日版本，适合需要版本一致性的应用。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"o3-mini-high","description":"OpenAI o3-mini高推理强度版本，模型会\"思考得更深入\"。处理复杂挑战时提供更高准确度，适合需要最佳精确度的场景，但需要更长的处理时间。ChatGPT付费用户可在模型选择器中使用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.15,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按次计费"]},{"model_name":"suno_act_stems","description":"Suno AI音轨分离功能（Stem Separation），将完整音乐分离为人声和乐器伴奏两个独立音轨。适合音乐制作、混音、卡拉OK制作、翻唱创作等专业音频处理场景。","tags":"音乐","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["C2","default","G1","G2","B1"],"supported_endpoint_types":["openai","Suno提交","Suno任务查询","Suno任务详情"],"billing_labels":["按次计费"]},{"model_name":"doubao-seedream-5-0","description":"字节跳动 Seedream 5.0 新一代图像生成与编辑模型，面向专业视觉创作，增强复杂构图、真实感、文字渲染和多图一致性。","tags":"图片","vendor_id":5,"created_time":1772252882,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","B1"],"supported_endpoint_types":["openai","image-generation","image-edits"],"billing_labels":["按次计费"]},{"model_name":"grok-video-3-pro","description":"支持10s视频","tags":"视频","vendor_id":8,"created_time":1770279671,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","openai-video"],"billing_labels":["按次计费"]},{"model_name":"gpt-5.3-chat-latest","description":"OpenAI GPT-5.3 对话模型，面向自然交互、复杂指令、代码辅助和工具调用优化，适合生产级聊天与智能体应用。该名称为动态更新别名，底层版本可能随服务商升级而变化。","tags":"文本","vendor_id":1,"created_time":1785210665,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["G2","C2","G1"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-5-20251101-thinking","description":"Claude Opus 4.5 是 Anthropic 于 2025 年 11 月推出的前沿推理模型，专为复杂的软件工程、智能体（Agent）工作流及长周期计算机操作而优化。该模型在代码编写、多步逻辑推理及工具调用方面表现出色，并具备极高的推理效率与上下文处理能力，被广泛应用于自动化研究、深度调试及企业级复杂任务处理。当前名称为按次计费别名，实际模型名称为 claude-opus-4-5-20251101-thinking。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 11 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["G2","B1","C2","default","G1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-4-32k-0613","description":"OpenAI GPT-4 系列高能力通用模型，擅长复杂指令遵循、逻辑推理、代码生成和专业文本处理，是早期 GPT-4 代旗舰模型。该名称固定到 2023 年 6 月 13 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":30,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["G1","G2","C2","default"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-5-20251101-thinking","description":"Claude Opus 4.5 是 Anthropic 于 2025 年 11 月推出的前沿推理模型，专为复杂的软件工程、智能体（Agent）工作流及长周期计算机操作而优化。该模型在代码编写、多步逻辑推理及工具调用方面表现出色，并具备极高的推理效率与上下文处理能力，被广泛应用于自动化研究、深度调试及企业级复杂任务处理。该版本启用增强思考模式，适合需要多步分析、复杂推理和代码规划的任务。该名称固定到 2025 年 11 月 1 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"gemini-2.5-pro-preview-06-05","description":"Google Gemini 2.5 Pro预览版（2025年6月5日版本），提供早期访问体验。具备最先进的推理能力，在编码、推理和多模态理解方面表现优异。支持原生音频输出、高级安全防护和Project Mariner的计算机使用功能。配备\"Deep Think\"实验性增强推理模式，专门针对复杂数学和编码问题进行优化。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["A1","B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-4o-mini-2024-07-18","description":"OpenAI GPT-4o 系列轻量多模态模型，支持文本与图像理解，具有速度快、成本低的特点，适合高并发对话和结构化处理。该名称固定到 2024 年 7 月 18 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-pro-high","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":6,"created_time":1771912459,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["B1","default"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"gemini-3.1-pro-preview-medium","description":"Google Gemini 3.1 Pro 高能力多模态推理模型，面向复杂编码、长上下文分析、工具调用和智能体任务，支持文本、图像、音频与视频理解。该条目将推理强度设为中档，以在速度、成本和回答质量之间提供固定配置。","tags":"文本","vendor_id":6,"created_time":1771912506,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["G1","G2","gemini-vertex","C2"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-sonnet-4-5-20250929-thinking","description":"Claude Sonnet 4.5增强思考版，具备可见的逐步推理能力。展示完整的思考过程，适合需要透明化决策逻辑的复杂任务场景。继承主模型的所有核心能力，在编码、数学推理和代理构建方面表现卓越。","tags":"文本","vendor_id":2,"created_time":1766657388,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["C2","default","G1","G2","A1","B1"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按量计费"]},{"model_name":"grok-4.5","description":"Grok 4.5是SpaceXAI（原xAI）于2026年7月发布的旗舰级大语言模型，基于1.5万亿参数的V9混合专家（MoE）架构打造，定位为Opus级别的编码与智能体专用模型，通过与Cursor联合引入数万亿级真实开发交互数据进行深度训练以显著强化代码生成、跨文件工程处理及长程AI Agent任务能力；该模型推理速度达约80 TPS，Token效率约为同类顶尖模型的2倍（在SWE Bench Pro任务中输出消耗仅为Opus 4.8的四分之一左右），API定价为每百万输入Token 2美元、输出6美元，具备50万（即将升级至100万）上下文窗口并支持多模态输入。","tags":"文本","vendor_id":8,"created_time":1784110218,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.25,"enable_groups":["B1","C2","codex渠道-gpt","default","G1","G2","A1","A2"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"qwen-vl-ocr-latest","description":"视觉理解文字识别模型最新版本，基于Qwen3-VL架构，显著改进了文档解析和文本定位能力。","tags":"文本","vendor_id":4,"created_time":1737782400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":1.666666666667,"enable_groups":["default","G1","G2","A2","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"glm-5.2","description":"GLM-5.2​ 是智谱 AI（Zhipu AI）于 2026年6月​ 发布并开源的新一代旗舰大语言模型，专为长程任务（Long-Horizon Tasks）与高级编程（Coding）场景设计。它在 Artificial Analysis 综合榜单取得 51 分，跻身全球模型前三，同时也是当前开源模型中的 SOTA（当前最优）水平。","tags":"文本","vendor_id":7,"created_time":1784109703,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"enable_groups":["C2","default","G1","G2","A1","A2","B1"],"supported_endpoint_types":["openai"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1-2025-04-14","description":"OpenAI GPT-4.1 旗舰非推理模型，针对编码、指令遵循和长上下文理解优化，适合软件开发、文档分析和工具调用。该名称固定到 2025 年 4 月 14 日版本，适合需要版本一致性的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1-mini","description":"OpenAI GPT-4.1 系列轻量模型，具备优秀的指令遵循、代码能力和长上下文理解，适合兼顾性能、速度与成本的应用。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-4.1-nano","description":"OpenAI GPT-4.1 系列最轻量模型，提供低延迟和低成本的长上下文处理，适合分类、抽取、自动补全及高吞吐任务。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1762411850,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"gpt-5-high","description":"OpenAI GPT-5 旗舰通用模型，具备强大的推理、编码、工具调用和多步骤任务执行能力，适合复杂知识工作与智能体应用。该条目将推理强度设为高档，以在速度、成本和回答质量之间提供固定配置。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","G1","G2","A1","B1","C2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]},{"model_name":"cs-claude-opus-4-1-20250805","description":"Anthropic Claude Opus 4.1 旗舰模型，重点增强真实代码库修改、长程智能体任务和复杂推理能力，适合高难度软件工程与研究。当前名称为按次计费别名，实际模型名称为 claude-opus-4-1-20250805。该名称固定到 2025 年 8 月 5 日版本，适合需要版本一致性的应用。","icon":"Claude.Color","tags":"文本","vendor_id":2,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gemini-3-pro-image-preview","description":"Google于2025年11月发布的图像生成模型（社区称Nano Banana Pro），支持最高4K分辨率和多轮对话式编辑。采用推理优先设计，在生成前进行规划，减少幻觉内容。内置搜索接地模块获取实时信息，支持文字转图像、文字渲染、交错生成等多种模式，专注于复杂、多轮、高保真的专业级生成任务。","tags":"图片","vendor_id":6,"created_time":1763690476,"quota_type":1,"model_ratio":0,"model_price":0.3,"owner_by":"","completion_ratio":0,"enable_groups":["B1","default","Gemini-image-G1"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按次计费"]},{"model_name":"seedance-2-0-4k-i2v","description":"支持 4K 分辨率。Seedance 2.0 标准视频生成模型，仅支持文生视频和图生视频，适合多镜头叙事、电影感镜头与音画同步创作。","tags":"视频","vendor_id":5,"created_time":1786853787,"quota_type":1,"model_ratio":0,"model_price":4.5,"model_price_type":"second","owner_by":"","completion_ratio":0,"enable_groups":["B1","default"],"supported_endpoint_types":["openai","OpenAI视频"],"billing_labels":["按次计费"]},{"model_name":"gemini-2.5-flash-lite","description":"Google推出的Gemini 2.5系列中最具成本效益和最快的AI模型，专为高吞吐量任务优化。相比之前版本具有更低的延迟，特别适用于翻译、分类等需要大规模处理的场景。保持了Gemini 2.5的核心能力，但在速度和成本方面更具优势，是需要快速响应和高并发处理的理想选择。","tags":"文本","vendor_id":6,"created_time":1762411850,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["A1","A2","default","G2","B1","C2","G1","gemini-vertex"],"supported_endpoint_types":["gemini","openai"],"billing_labels":["按量计费"]},{"model_name":"claude-opus-4-8-c","description":"Claude-Opus-4-8​ 是 Anthropic 于 2026年5月28日​ 发布的 Opus 系列旗舰大语言模型，也是当时正式发布的最强 Opus 层级模型，专为复杂推理、长周期自主智能体（Agent）编程与企业级知识工作设计，在编码、代理任务与判断可靠性上较 Opus 4.7 有显著迭代提升，默认支持 1M Token​ 上下文窗口与 128K​ 最大输出，并搭载优化后的自适应思考（Adaptive Thinking）与 effort 调节机制以平衡推理深度与成本。","tags":"文本","vendor_id":2,"created_time":1780298897,"quota_type":1,"model_ratio":0,"model_price":0.4,"owner_by":"","completion_ratio":0,"enable_groups":["default","G1","G2","B1","C2"],"supported_endpoint_types":["openai","anthropic"],"billing_labels":["按次计费"]},{"model_name":"gpt-image-1.5","description":"OpenAI于2025年12月发布的图像生成服务升级版本，比GPT Image 1快4倍，提示词对齐度达91.2%。支持工作室级别的图像质量、增强的图像保留（面部相似度、光线、构图、色调）、精准文本渲染（排版、标志、代码）、精确局部编辑。","tags":"图片","vendor_id":1,"created_time":1766666400,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["B1","C2","default","G1","G2"],"supported_endpoint_types":["image-generation","openai","image-edits"],"billing_labels":["按次计费"]},{"model_name":"o3-mini-medium","description":"OpenAI o3-mini中等推理强度版本，ChatGPT中的默认设置。在速度和准确度之间提供最佳平衡，适合大多数常规推理任务和日常使用场景。","icon":"OpenAI.Color","tags":"文本","vendor_id":1,"created_time":1766666400,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["C2","G1","G2"],"supported_endpoint_types":["openai","openai-response"],"billing_labels":["按量计费"]}],"group_ratio":{"A1":0.5,"A2":0.9,"B1":0.9,"C2":3.5,"Claude Max":1.8,"Claude-Aws":4,"Claude-G1":4.5,"G1":5,"G2":2.5,"Gemini-image-G1":0.8,"auto":1.5,"codex渠道-gpt":0.28,"default":1.2,"gemini-vertex":3.6},"pricing_version":"c576c64d115cf2ed5711aff7006a5d7ae1b7f12e6a5bbd858b36d51fa0601112","success":true,"supported_endpoint":{"Anthropic":{"path":"/v1/messages","method":"POST"},"Gemini":{"path":"/v1beta/models/{model}:generateContent","method":"POST"},"Midjourney Fast任务列表":{"path":"/fast/mj/task/list-by-condition","method":"POST"},"Midjourney Fast任务详情":{"path":"/fast/mj/task/{id}/fetch","method":"GET"},"Midjourney Fast动作放大（Plus）":{"path":"/fast/mj/submit/action","method":"POST"},"Midjourney Fast动作（Plus）":{"path":"/fast/mj/submit/action","method":"POST"},"Midjourney Fast变更":{"path":"/fast/mj/submit/change","method":"POST"},"Midjourney Fast图片代理":{"path":"/fast/mj/image/{id}","method":"GET"},"Midjourney Fast图片种子":{"path":"/fast/mj/task/{id}/image-seed","method":"GET"},"Midjourney Fast图生文":{"path":"/fast/mj/submit/describe","method":"POST"},"Midjourney Fast局部编辑":{"path":"/fast/mj/submit/modal","method":"POST"},"Midjourney Fast提示词缩短":{"path":"/fast/mj/submit/shorten","method":"POST"},"Midjourney Fast混图":{"path":"/fast/mj/submit/blend","method":"POST"},"Midjourney Fast简化变更":{"path":"/fast/mj/submit/simple-change","method":"POST"},"Midjourney Fast绘图":{"path":"/fast/mj/submit/imagine","method":"POST"},"Midjourney Fast编辑":{"path":"/fast/mj/submit/edits","method":"POST"},"Midjourney Relax任务列表":{"path":"/relax/mj/task/list-by-condition","method":"POST"},"Midjourney Relax任务详情":{"path":"/relax/mj/task/{id}/fetch","method":"GET"},"Midjourney Relax动作放大（Plus）":{"path":"/relax/mj/submit/action","method":"POST"},"Midjourney Relax动作（Plus）":{"path":"/relax/mj/submit/action","method":"POST"},"Midjourney Relax变更":{"path":"/relax/mj/submit/change","method":"POST"},"Midjourney Relax图片代理":{"path":"/relax/mj/image/{id}","method":"GET"},"Midjourney Relax图片种子":{"path":"/relax/mj/task/{id}/image-seed","method":"GET"},"Midjourney Relax图生文":{"path":"/relax/mj/submit/describe","method":"POST"},"Midjourney Relax局部编辑":{"path":"/relax/mj/submit/modal","method":"POST"},"Midjourney Relax提示词缩短":{"path":"/relax/mj/submit/shorten","method":"POST"},"Midjourney Relax混图":{"path":"/relax/mj/submit/blend","method":"POST"},"Midjourney Relax简化变更":{"path":"/relax/mj/submit/simple-change","method":"POST"},"Midjourney Relax绘图":{"path":"/relax/mj/submit/imagine","method":"POST"},"Midjourney Relax编辑":{"path":"/relax/mj/submit/edits","method":"POST"},"Midjourney任务列表":{"path":"/mj/task/list-by-condition","method":"POST"},"Midjourney任务详情":{"path":"/mj/task/{id}/fetch","method":"GET"},"Midjourney动作（Plus）":{"path":"/mj/submit/action","method":"POST"},"Midjourney变更":{"path":"/mj/submit/change","method":"POST"},"Midjourney图片上传":{"path":"/mj/submit/upload-discord-images","method":"POST"},"Midjourney图片代理":{"path":"/mj/image/{id}","method":"GET"},"Midjourney图片种子":{"path":"/mj/task/{id}/image-seed","method":"GET"},"Midjourney图生文":{"path":"/mj/submit/describe","method":"POST"},"Midjourney局部重绘":{"path":"/mj/submit/modal","method":"POST"},"Midjourney提示词缩短":{"path":"/mj/submit/shorten","method":"POST"},"Midjourney混图":{"path":"/mj/submit/blend","method":"POST"},"Midjourney简化变更":{"path":"/mj/submit/simple-change","method":"POST"},"Midjourney绘图":{"path":"/mj/submit/imagine","method":"POST"},"OpenAI对话":{"path":"/v1/chat/completions","method":"POST"},"OpenAI对话（响应式）":{"path":"/v1/responses","method":"POST"},"OpenAI改图":{"path":"/v1/images/edits","method":"POST"},"OpenAI文本转语音":{"path":"/v1/audio/speech","method":"POST"},"OpenAI生图":{"path":"/v1/images/generations","method":"POST"},"OpenAI视频":{"path":"/v1/videos","method":"POST"},"OpenAI语音转文本":{"path":"/v1/audio/transcriptions","method":"POST"},"Suno Open任务查询":{"path":"/open/suno/fetch","method":"POST"},"Suno Open任务详情":{"path":"/open/suno/fetch/{id}","method":"GET"},"Suno Open提交":{"path":"/open/suno/submit/music","method":"POST"},"Suno任务查询":{"path":"/suno/fetch","method":"POST"},"Suno任务详情":{"path":"/suno/fetch/{id}","method":"GET"},"Suno提交":{"path":"/suno/submit/music","method":"POST"},"anthropic":{"path":"/v1/messages","method":"POST"},"audio-speech":{"path":"/v1/audio/speech","method":"POST"},"audio-transcriptions":{"path":"/v1/audio/transcriptions","method":"POST"},"audio-translations":{"path":"/v1/audio/translations","method":"POST"},"embeddings":{"path":"/v1/embeddings","method":"POST"},"gemini":{"path":"/v1beta/models/{model}:generateContent","method":"POST"},"image-edits":{"path":"/v1/images/edits","method":"POST"},"image-generation":{"path":"/v1/images/generations","method":"POST"},"openai":{"path":"/v1/chat/completions","method":"POST"},"openai-realtime":{"path":"/v1/realtime","method":"GET"},"openai-response":{"path":"/v1/responses","method":"POST"},"openai-response-compact":{"path":"/v1/responses/compact","method":"POST"},"openai-video":{"path":"/v1/videos","method":"POST"}},"updated":true,"usable_group":{"A1":"🔥限时特价-质量一般","A2":"国产模型推荐分组","B1":"🔥混合特价-按次模型更划算","C2":"🚀官转直连-企业级服务","Claude Max":"Claude Code MAX号池，含防封注入","Claude-Aws":"Aws优质渠道","Claude-G1":"Claude官转","G1":"🚀官转渠道-企业首选","G2":"🚀优质渠道-高性价比","Gemini-image-G1":"Gemini香蕉模型限时特价分组","auto":"系统测试，不要选择使用","codex渠道-gpt":"🔥限时特价-模型较少，详见广场","default":"默认分组-支持更多模型，质量一般","gemini-vertex":"Gemini高质量渠道"},"vendors":[{"id":2,"name":"Anthropic","icon":"Claude.Color"},{"id":4,"name":"阿里巴巴","icon":"Qwen.Color"},{"id":6,"name":"Google","icon":"Gemini.Color"},{"id":33,"name":"其它"},{"id":34,"name":"MiniMax","icon":"Minimax.Color"},{"id":3,"name":"DeepSeek","icon":"DeepSeek.Color"},{"id":8,"name":"xAI","icon":"XAI"},{"id":11,"name":"讯飞","icon":"Spark.Color"},{"id":27,"name":"即梦","icon":"Jimeng.Color"},{"id":39,"name":"Mistral","icon":"Mistral.Color"},{"id":40,"name":"零一万物","icon":"Yi.Color"},{"id":45,"name":"HappyHorse"},{"id":48,"name":"Xiaomi MiMo","description":"Xiaomi MiMo 是小米公司自研的推理大模型系列‌，2025 年 4 月首次开源，目前最新可用版本为 MiMo-V2.5 系列，提供在线对话、API 调用和本地部署等多种使用方式。","icon":"XiaomiMiMo.Avatar size={64} shape={'square'} "},{"id":5,"name":"字节跳动","icon":"Doubao.Color"},{"id":7,"name":"智谱","icon":"Zhipu.Color"},{"id":9,"name":"Moonshot","icon":"Moonshot"},{"id":35,"name":"KwaiKAT (KAT-Coder)","icon":"KwaiKAT"},{"id":36,"name":"百度","icon":"Wenxin.Color"},{"id":37,"name":"腾讯","icon":"Hunyuan.Color"},{"id":1,"name":"OpenAI","icon":"OpenAI"},{"id":10,"name":"快手","icon":"Kling.Color"},{"id":26,"name":"Midjourney","description":"支持Midjourney所有操作 Mj V7、局部重绘、 支持切换绘画模式 fast、turbo、relax,可自行接入，支持Midjourney proxy plus以及Midjourney proxy接口协议快速接入，已集成了图片国内反代+Discord国内反代+中文翻译接口，高并发场景并发。","icon":"Midjourney.Color"},{"id":38,"name":"Meta","icon":"Ollama"},{"id":46,"name":"Cloudflare","icon":"Cloudflare.Color"}]}