{"auto_groups":[],"data":[{"model_name":"claude-sonnet-4-6-thinking","description":"Claude Sonnet 4.6 是 Anthropic 在 2026 年 2 月中旬推出的最新 Sonnet 系列模型，被定位为“更便宜、更快、但能力更接近旗舰 Opus”的通用模型；官方与媒体解读普遍强调它在编程能力、计算机使用/办公自动化、长上下文推理与智能体规划等方面有明显升级","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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":["AWS正价"],"supported_endpoint_types":["anthropic","openai"],"pricing_version":"5a90f2b86c08bd983a9a2e6d66c255f4eaef9c4bc934386d2b6ae84ef0ff1f1f"},{"model_name":"claude-haiku-4-5","vendor_id":19,"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":["CC MAX"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-5.5","description":"GPT-5.5 是 OpenAI 于 2026 年 4 月 23 日正式推出的最新旗舰模型，是自 GPT-4.5 以来首次完全重新训练的基础模型（内部代号“Spud”）。它专为真实工作场景设计，在智能体编码、计算机使用、知识工作和科学研究上实现显著跃升，能更快理解用户意图，自主规划并执行多步骤复杂任务（如调试代码、操作软件、分析数据、生成报告），同时 token 效率更高、响应更快，在 Terminal-Bench、OSWorld、GeneBench 等基准测试中全面超越 GPT-5.4","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["codex","优质GPT"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"(len \u003c= 272000 ? tier(\"standard\", p * 5 + c * 30 + cr * 0.5) : tier(\"long_context\", p * 10 + c * 60 + cr * 1)) * (param(\"service_tier\") == \"priority\" \u0026\u0026 len \u003c= 272000 ? 2.5 : 1) * (param(\"service_tier\") == \"fast\" \u0026\u0026 len \u003c= 272000 ? 2.5 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"gpt-6-astra","description":"GPT-6 Astra 是 OpenAI 面向高难度端到端任务推出的新一代旗舰模型，重点强化了复杂推理、软件工程、Computer Use、深度研究、网页操作以及长流程 Agent 智能体能力。它能够自主规划并执行多步骤任务，在代码开发、专业工作、科研分析和自动化场景中拥有更强的持续执行能力。GPT-6 Astra 支持约 105 万 Token 上下文窗口和最高 128K Token 输出，并提供从 Low 到 Max 的多档推理强度，适合大型代码库开发、复杂研究、文档处理、企业级 Agent 和高难度自动化任务。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["优质GPT","codex"],"supported_endpoint_types":["openai"],"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 * 2 + cc * 25)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"claude-opus-5","description":"Claude Opus 5（API 标识 claude-opus-5）是 Anthropic 于 2026 年 7 月 24 日发布的最新 Opus 系列模型，定位为面向复杂 agentic 编程和企业级工作的日常主力模型。它在编程、长程多步骤任务、知识工作等评测（如 Frontier-Bench、CursorBench、OSWorld 等）上取得显著提升，接近更高阶的 Claude Fable 5 的整体能力，但价格仅为后者一半（输入/输出 $5/$25 每百万 tokens，与前代 Opus 4.8 持平），同时具备 100 万 tokens 上下文窗口、最高 12.8 万 tokens 输出，以及截至 2026 年 5 月的知识截止时间","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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按量","aws","AWS正价","aws限时特价","CC MAX","Claude Code"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"[V]gemini-3-pro-image-preview","icon":"Gemini.Avatar.type={'platform'}","vendor_id":15,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":60,"enable_groups":["Banana_GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-5.6-luna","description":"gpt-5.6-luna 是 GPT-5.6 系列中成本最低、速度最快的版本。它针对高并发、高吞吐量的场景优化，在保持一定智能水平的同时大幅降低了使用门槛（短上下文输入 $1 / 输出 $6）。适合大规模内容处理、客服机器人、数据标注等对单次调用成本敏感且不需要最强推理能力的应用，是高频低价场景的实用选择。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["codex","优质GPT"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"(len \u003c= 272000 ? tier(\"standard\", p * 0.2 + c * 1.2 + cr * 0.02 + cc * 0.25) : tier(\"long_context\", p * 0.4 + c * 1.8 + cr * 0.04 + cc * 0.5)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"gemini-3.1-pro-preview","description":"gemini-3.1-pro-preview 是 Google 推出的一款针对软件工程任务和可用性进行深度优化的预览版大语言模型，它不仅具备 100 万 token 的长上下文处理能力，还特别针对复杂的智能体（Agent）工作流进行了强化，使其能够在真实开发场景中更精确地优先调用自定义工具（如搜索代码或执行命令行操作）并极其可靠地完成多步骤的复杂任务","icon":"Gemini.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.07,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini","Gemini混合","Gemini特价"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3.8-flash","description":"Gemini 3.8 Flash 是 Google 最新一代高性能 Flash 模型，也是目前 Gemini Flash 系列中智能能力最强的版本。相比 3.7 Flash，它进一步提升了复杂推理、长流程软件开发、Agent 智能体和多步骤任务处理能力，同时继续兼顾较快的响应速度与较低的调用成本。模型支持文本、图片、音频、视频和 PDF 等多模态输入，拥有约 100 万 Token 上下文窗口和最高 64K Token 输出，并支持代码执行、函数调用、联网搜索、Computer Use 以及可调节的 Thinking 推理强度，适合编程开发、复杂任务自动化、企业应用和大规模 API 调用等场景。","icon":"Gemini.Avatar.type={'platform'}","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.02,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini_flash","Gemini混合","Gemini特价","Gemini"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-image-2.5","description":"GPT-Image-2.5 是 OpenAI 最新一代图像生成与编辑模型，相比 GPT-Image-2，重点提升了画面清晰度、人物与参考图一致性、自然光影和纹理表现、复杂指令理解以及局部精准编辑能力，在多轮修改中也能更好地保持人物、构图和风格不变，同时生成延迟最高可降低约 50%。","tags":"生图","vendor_id":2,"quota_type":1,"model_ratio":0,"model_price":0.25,"owner_by":"","completion_ratio":0,"enable_groups":["优质GPT"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-4-7","description":"Claude Opus 4.7 是 Anthropic 于 2026 年 4 月 16 日正式发布的最新旗舰级大型语言模型，作为公司当前最强大的通用可用模型，它在复杂推理、agentic coding（智能体编码）、多步任务处理以及专业知识工作领域实现了显著跃升，较前代 Opus 4.6 模型在可靠性和一致性方面取得突破性进步。该模型支持高达 100 万 token 的上下文窗口，具备出色的视觉理解能力，能够高效处理图像输入，并适用于高要求的软件工程、长期 agentic 工作流以及企业级复杂任务场景，同时严格遵循 Anthropic 的 ASL-3 安全标准，确保在提供前沿智能的同时维持高度的可控性与安全性，是专业用户和开发团队处理高难度工作的首选模型。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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":["AWS正价","aws限时特价","CC MAX","Claude Code","Claude按量","aws"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-sonnet-4-5-20250929","description":"claude-sonnet-4-5-20250929 是2025年9月推出的新一代模型，被誉为全球顶尖的编码AI，其核心突破在于能够自主规划并执行超长周期的复杂任务（例如持续编码数十小时），在多项性能基准测试中创造了新纪录。","icon":"Claude.Avatar.type={'platform'}","vendor_id":19,"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":["aws","aws限时特价","CC MAX","Claude Code","Claude按量"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-sonnet-4-5-20250929-thinking","description":"claude-sonnet-4-5-20250929-thinking 为 Sonnet-4 的中期重要升级版（即“.5”版本）。重点优化和增强特定能力。此版本可能在继承 Sonnet-4 的所有功能基础上，显著提升了推理规划的准确性、工具调用的稳定性和执行效率，或是引入了全新的模态（如视频理解）支持。后缀 “-thinking” 依然是其核心卖点，日期 “20250929” 则标志着这个增强版本的成熟时间点，代表了 Anthropic 在智能体技术上的持续快速迭代。","icon":"Claude.Avatar.type={'platform'}","vendor_id":19,"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":["aws","aws限时特价","Claude按量"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-6-sol","description":"GPT-6 Sol是OpenAI 于 2026 年 9 月推出的 GPT-6 系列主力推理模型，定位于高强度编程、Agent 工作流和复杂知识任务，在能力、速度和成本之间取得平衡。它支持最高约 105 万 Token 上下文、12.8 万 Token 最大输出以及多档 reasoning effort，特别适合代码开发、调试、工具调用、长上下文分析和自动化工作流。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"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":["codex","优质GPT"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"(len \u003c= 272000 ? tier(\"standard\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5) : tier(\"long_context\", p * 4 + c * 15 + cr * 0.4 + cc * 5)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"[V]gemini-3.1-pro-preview","description":"gemini-3.1-pro-preview 是 Google 推出的一款针对软件工程任务和可用性进行深度优化的预览版大语言模型，它不仅具备 100 万 token 的长上下文处理能力，还特别针对复杂的智能体（Agent）工作流进行了强化，使其能够在真实开发场景中更精确地优先调用自定义工具（如搜索代码或执行命令行操作）并极其可靠地完成多步骤的复杂任务","icon":"Gemini.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":15,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini","纯血GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-opus-5-5","description":"Claude Opus 5.5是Anthropic 于 2026 年 9 月推出的新一代旗舰模型，重点面向复杂编程、Agent、多步骤推理和专业知识工作。相比上一代 Opus 5，它进一步提升了大型代码库处理、长流程自主执行和复杂任务稳定性，同时运行成本有所下降，适合软件开发、代码重构、研究分析以及需要持续执行大量步骤的高难度任务","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.05,"create_cache_ratio":1.25,"enable_groups":["AWS正价","aws限时特价","CC MAX","Claude Code","Claude按量","aws"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"[V]gemini-3-pro-preview","description":"Gemini 3 Pro Preview 是谷歌新一代的旗舰级多模态大模型，其核心优势在于强大的通用推理能力、复杂的多模态理解以及卓越的编程与数学能力。它被设计为一个统一的平台，能够原生处理和理解文本、图像、音频、视频和代码等多种信息形式，并支持高达104万tokens的上下文窗口，使其能够处理极其冗长的文档或代码库。该模型在多项学术和专业基准测试中取得了突破性成绩，展示了其博士级别的推理水平。此外，它还引入了 Deep Think模式，通过更长时间的深度思考，专门用于解决最复杂的逻辑推理和长周期规划任务。","icon":"Gemini.Avatar.type={'platform'}","tags":"对话,多模态","vendor_id":15,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["按量Gemini","纯血GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"[V]gemini-3.5-flash","description":"Gemini 3.5 Flash 是 Google 于 2026 年 5 月发布的 Gemini 系列最新 Flash 模型，定位为目前 Flash 家族中智能水平最高的版本。它试图在保持 Flash 一贯的响应速度和较低延迟的同时，提供接近 Pro 模型的编码、推理和代理式任务能力，支持文本、图像、视频、音频等多模态输入，上下文窗口最高可达 100 万 tokens，并内置了较强的 Google Search grounding 功能","icon":"Gemini.Avatar.type={'platform'}","tags":"代码","vendor_id":15,"quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-2.5-flash","description":"Gemini 2.5 Flash在性价比方面表现出色的模型，可提供全面的功能。2.5 Flash 最适合大规模处理、低延迟、需要思考的高数据量任务，以及代理应用场景。","icon":"Gemini.Avatar.type={'platform'}","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.015,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini混合","Gemini特价","Gemini","Gemini_flash"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"[V]gemini-2.5-pro","description":"[vertex渠道]Gemini 2.5 Pro 是我们最先进的思考型模型，能够推理代码、数学和 STEM 领域的复杂问题，还能使用长上下文分析大型数据集、代码库和文档","icon":"Gemini.Avatar.type={'platform'}","tags":"对话","vendor_id":15,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["按量Gemini","纯血GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-2.5-flash-image","description":"Gemini 2.5 Flash Image （nano banana）是一款最新、最快、最高效的原生多模态模型，可让您以对话方式生成和编辑图像。 该模型具备多图像融合能力，能够理解并合并多个输入图像，例如将一个物体放入一个场景，或用一种配色方案或纹理重新设计一个房间的风格。 此外，它还支持基于提示的图像编辑，可以通过自然语言指令进行精确的局部编辑，例如模糊图像背景、去除T恤上的污渍或更改人物的姿势。 得益于Gemini的世界知识，该模型在图像生成方面不仅限于美学，还具备了对现实世界的深入语义理解。","icon":"Gemini.Avatar.type={'platform'}","tags":"生图","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["Banana增稳"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-opus-4-6","description":"Claude-opus-4-6 是 Anthropic 最新发布的一款 高级大语言模型（LLM），它在 Opus 系列基础上做了重要升级，特别强化了 复杂推理、长期规划、编码与知识工作能力，并支持超大的 百万 token 上下文窗口（beta），让模型能处理更长、更复杂的任务。相比前代，它在编程、调试、法律/金融分析等专业应用上表现更强，并减少了用户反复提示的需要，还展示了出色的安全与漏洞检测能力，是面向生产力工具和企业级知识工作的先进 AI 平台。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,工具调用","vendor_id":19,"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":["aws","AWS正价","aws限时特价","CC MAX","Claude Code","Claude按量"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-opus-4-6-thinking","description":"Claude-opus-4-6 是 Anthropic 最新发布的一款 高级大语言模型（LLM），它在 Opus 系列基础上做了重要升级，特别强化了 复杂推理、长期规划、编码与知识工作能力，并支持超大的 百万 token 上下文窗口（beta），让模型能处理更长、更复杂的任务。相比前代，它在编程、调试、法律/金融分析等专业应用上表现更强，并减少了用户反复提示的需要，还展示了出色的安全与漏洞检测能力，是面向生产力工具和企业级知识工作的先进 AI 平台。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,思考,工具调用","vendor_id":19,"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":["AWS正价"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gemini-3-pro-image-preview","description":"Gemini 3 Pro Image Preview 则是一个专注于视觉内容创作的专业模型，其核心功能是生成高达4K分辨率的高质量图像，并支持对话式的多轮图像编辑。该模型的一大技术亮点是将推理能力融入图像生成过程，这不仅能提升图像的细节质量和对长文本描述的渲染准确度，还能显著增强生成内容的事实准确性。它旨在处理高难度的图像生成任务，并通过与用户的对话，持续对图像进行迭代和编辑，从而实现高效的创意工作流。","icon":"Gemini.Avatar.type={'platform'}","tags":"生图","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.35,"owner_by":"","completion_ratio":0,"enable_groups":["Banana","Banana增稳"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-image-2","description":"GPT Image 2 是 OpenAI 开发的最新图像生成模型，于 2026 年 4 月正式发布。它作为 GPT Image 系列的最新版本，取代了之前的 DALL·E 模型，采用自回归架构，直接集成于 ChatGPT 和 OpenAI API 中。该模型在文本渲染、照片级真实感、复杂指令遵循以及图像编辑等方面表现出色，能够根据自然语言描述生成高分辨率（支持高达 4K）、细节丰富且上下文准确的图像，特别擅长处理文字内容、UI 界面和产品摄影等场景","icon":"OpenAI.Avatar.type={'platform'}","tags":"生图","vendor_id":2,"quota_type":1,"model_ratio":0,"model_price":0.2,"owner_by":"","completion_ratio":0,"enable_groups":["优质GPT"],"supported_endpoint_types":["openai","image"]},{"model_name":"[V]gemini-2.5-flash","description":"[vertex渠道]Gemini 2.5 Flash在性价比方面表现出色的模型，可提供全面的功能。2.5 Flash 最适合大规模处理、低延迟、需要思考的高数据量任务，以及代理应用场景。\n\n","icon":"Gemini.Avatar.type={'platform'}","tags":"对话","vendor_id":15,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333333333333,"cache_ratio":0.1,"enable_groups":["按量Gemini","纯血GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"[V]gemini-3-flash-preview","vendor_id":15,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini","纯血GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"[V]gemini-3.8-flash","vendor_id":15,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-fable-5","description":"Claude Fable 5 是 Anthropic 于 2026 年 6 月 9 日发布的模型，是该公司首个广泛向公众开放的 Mythos 级别模型。它与仅限受信合作伙伴使用的 Claude Mythos 5 共享相同底层能力，主要面向复杂的长期推理、自主代理工作、软件工程和知识密集型任务，支持默认 100 万 token 上下文窗口及最高 12.8 万 token 输出，定价为每百万输入 token 10 美元、输出 50 美元。相较 Mythos 5，Fable 5 版本增加了安全分类器，可在涉及网络安全、生物等高风险领域拒绝请求，以实现更广泛的安全部署","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用,旗舰模型","vendor_id":19,"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 Code","Claude按量","aws","AWS正价","aws限时特价","CC MAX"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-sonnet-5-5","description":"Claude Sonnet 5.5是 Anthropic 于 2026 年 9 月发布的 Claude 5.5 系列主力均衡型模型，相比上一代 Sonnet 5 在速度、成本和实际任务能力上都有明显提升，官方定位更偏向日常开发、Bug 修复、Agent 工作流以及高质量文档、表格和演示文稿制作，同时保持较强的视觉理解和设计能力。它可以看作 Opus 5.5 之下更强调速度与性价比的选择，尤其适合需要大量调用但又不想明显牺牲代码和生产力表现的场景。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["aws","AWS正价","aws限时特价","CC MAX","Claude Code","Claude按量"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5 + cc1h * 4)"},{"model_name":"gemini-3.1-flash-image-preview","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.18,"owner_by":"","completion_ratio":0,"enable_groups":["Banana增稳","Banana"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"[V]gemini-3.1-flash-image-preview","icon":"Gemini.Avatar.type={'platform'}","vendor_id":15,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":120,"enable_groups":["Banana_GCP"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"codex-auto-review","description":"codex-auto-review 是 Codex（OpenAI 的 AI 编程代理工具）中 Auto-review 自动审阅功能 专用的 reviewer 模型。它不是用来直接写代码的，而是作为独立的审阅代理，在主代理即将执行可能有风险的操作（如写文件、运行命令、访问网络等）时，自动判断该操作是否安全并决定是否批准。开启 approvals_reviewer = \"auto_review\" 后，Codex 就能在长任务中减少频繁的人工确认，同时通过这个模型来把控安全边界。","icon":"OpenAI.Avatar.type={'platform'}","tags":"审查","vendor_id":2,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["codex","优质GPT"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.6-sol","description":"gpt-5.6-sol 是 GPT-5.6 系列中的旗舰模型，定位为当前 OpenAI 智能水平最高的版本。它在复杂推理、长链条规划、编程、代理任务以及专业领域（如网络安全、科学分析）上表现最强，适合对输出质量要求极高的场景。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["codex","优质GPT"],"supported_endpoint_types":["openai","openai-response"],"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 * 60 + cr * 1 + cc * 12.5)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"claude-haiku-4-5-20251001","description":"claude-haiku-4-5-20251001 是 Anthropic 轻量级系列的重大版本更新，定位为市场上速度最快、成本效益最高的模型之一。版本号 “4.5” 暗示这是一个性能大幅跃升的版本，可能在响应速度、上下文处理效率和基础推理能力上超越了前代 Haiku 模型。它的主要应用场景是处理高并发、低延迟的实时交互任务，如实时翻译、内容审核和智能客服等。虽然名称中未包含 “-thinking”，但预计它会集成更可靠的基础工具调用功能，以满足轻量级自动化任务的需求。","icon":"Claude.Avatar.type={'platform'}","vendor_id":19,"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":["Claude按量","aws","AWS正价","aws限时特价","CC MAX","Claude Code"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-opus-4-5","vendor_id":19,"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":["CC MAX"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gemini-3.5-flash","description":"gemini-3.5-flash 是 Google 最新的 Flash 系列高效率模型，定位是在尽量保持低延迟、低成本的同时，提供接近旗舰大模型的推理、编程和多步骤任务能力；官方说明它已 GA 稳定可用于生产环境，特别适合 agent 工作流、长时间任务、复杂代码迭代、多模态输入和大规模并发场景。简单说，它不是最“极限智商”的 Pro 旗舰，而是更适合中转站、应用接入、自动化代理、代码助手这类需要速度、成本、稳定性、能力均衡的模型。","icon":"Gemini.Avatar.type={'platform'}","tags":"对话,多模态,代码","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini特价","Gemini","Gemini_flash","Gemini混合"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-opus-4-8","description":"Claude Opus 4.8 是 Anthropic 公司于 2026 年 5 月 28 日发布的最新旗舰 AI 模型，为 Claude Opus 系列的升级版本。它是 Anthropic 当前最强大的通用可用模型，在编码、AI 代理任务以及专业知识工作中表现出显著提升，尤其擅长复杂推理、长时间自主协作以及大规模代码处理，支持高达 100 万 token 的上下文窗口，并具备更高的自主性和一致性。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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 Code","Claude按量","aws","AWS正价","aws限时特价","CC MAX"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-5.6-terra","description":"gpt-5.6-terra 是平衡型模型，在智能水平和成本之间做了较好的取舍。其性能接近上一代旗舰 GPT-5.5，但价格仅为 Sol 的一半（短上下文输入 $2.5 / 输出 $15），适合大多数生产环境下的日常高价值任务。无论是内容生成、代码辅助还是多步骤工作流，它都能提供可靠的性价比，是目前多数开发者首选的中间档位。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["codex","优质GPT"],"supported_endpoint_types":["openai","openai-response"],"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)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"gpt-6.1-sol","description":"GPT-6.1 Sol是 OpenAI 于 2026 年 9 月推出的新一代 Sol 系列主力模型，是 GPT-6 Sol 的升级版，重点强化了复杂编程、代码调试、长流程 Agent、计算机操作以及专业文档分析等能力，在多项高难度任务上已经接近旗舰 GPT-6 Astra，同时保持明显更低的运行成本。它支持约 105 万 Token 上下文、12.8 万 Token 最大输出，比较适合大型代码项目、复杂推理、文档处理和需要多步骤执行的自动化工作流。","icon":"OpenAI.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["优质GPT","codex"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"(len \u003c= 272000 ? tier(\"standard\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5) : tier(\"long_context\", p * 4 + c * 15 + cr * 0.4 + cc * 5)) * (param(\"service_tier\") == \"priority\" ? 2 : 1) * (param(\"service_tier\") == \"fast\" ? 2 : 1) * (param(\"service_tier\") == \"flex\" ? 0.5 : 1)"},{"model_name":"[V]gemini-3.7-flash","vendor_id":15,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-sonnet-5","description":"Claude Sonnet 5 是 Anthropic 于 2026 年最新发布的 Sonnet 系列模型，定位为兼顾高性能与实用成本的“最强 Sonnet”。它在编码、长时程代理、工具调用和计算机使用等任务上接近 Opus 旗舰水平，同时保持 Sonnet 系列的速度优势，支持 100 万 token 上下文窗口。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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":["CC MAX","Claude Code","Claude按量","aws","AWS正价","aws限时特价"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-opus-4-5-20251101","description":"Claude Opus 4.5是Anthropic于2025年11月发布的最强大模型，在软件工程等任务上展现了顶尖性能。它不仅在处理复杂任务时拥有更强的判断力和推理能力，还实现了大幅降价，成本比前代降低约三分之二，同时在响应效率上也得到了显著提升，是目前综合能力极强且性价比更高的一款尖端AI模型。","icon":"Claude.Avatar.type={'platform'}","tags":"对话","vendor_id":19,"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":["aws","aws限时特价","CC MAX","Claude Code","Claude按量"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-sonnet-4-6","description":"Claude Sonnet 4.6 是 Anthropic 在 2026 年 2 月中旬推出的最新 Sonnet 系列模型，被定位为“更便宜、更快、但能力更接近旗舰 Opus”的通用模型；官方与媒体解读普遍强调它在编程能力、计算机使用/办公自动化、长上下文推理与智能体规划等方面有明显升级","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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":["aws","AWS正价","aws限时特价","CC MAX","Claude Code","Claude按量"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"[V]gemini-3.6-flash","vendor_id":15,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["按量Gemini","特价按量Gemini"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-fable-5-1","description":"Claude Fable 5.1 是 Anthropic 于 2026 年 9 月推出的高端旗舰模型，主要面向高难度推理、复杂编程以及长周期 Agent 智能体任务。相比 Fable 5，它进一步强化了长时间自主编程、多步骤研究、文档与数据分析等能力，能够持续规划、调用工具、处理失败并完成跨应用的复杂工作流。模型支持文本和图像输入，拥有 100 万 Token 上下文窗口和最高 128K Token 输出，采用始终开启的自适应 Thinking 推理机制，尤其适合大型代码库开发、深度研究、企业自动化和需要长时间连续执行的复杂任务","icon":"Claude.Avatar.type={'platform'}","tags":"对话,代码,工具调用","vendor_id":19,"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":["aws限时特价","CC MAX","Claude Code","Claude按量","aws","AWS正价"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gemini-3-flash-preview","description":"Gemini 3 Flash 是 Google 于 2025 年 12 月最新发布的高效率旗舰模型，专为在极致速度和低成本下提供前沿级智能而设计。\n​作为 Gemini 3 系列的轻量化版本，它不仅继承了百万级（1M）的长上下文处理能力，更在视觉空间推理和**代理编程（Agentic Coding）**上实现了巨大飞跃，在许多代码和多模态任务中拥有媲美甚至超越此前 Pro 级模型的表现。简单来说，它是目前综合性价比极高、响应极快的主力模型，非常适合需要实时交互、复杂指令跟随以及大规模 API 调用的业务场景。","icon":"Gemini.Avatar.type={'platform'}","tags":"对话","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.01,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini","Gemini_flash","Gemini混合","Gemini特价"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3.6-flash","description":"Gemini 3.6 Flash 是 Google 于 2026 年 7 月 21 日发布的 Gemini 3 系列模型，作为 3.5 Flash 的后续版本，被定位为编码、知识处理和多模态任务的日常主力模型。它在输出 token 效率上有所优化，据称比前代减少约 17%，并在部分编码基准上报告了性能提升","icon":"Gemini","tags":"对话,代码,工具调用","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.035,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini","Gemini混合","Gemini特价"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3.7-flash","description":"Gemini 3.7 Flash 是 Google 推出的新一代高性能 Flash 模型，主打**速度、成本与智能能力的平衡**，尤其针对代码生成、复杂推理、智能体（Agent）工作流和多步骤任务进行了强化。它支持文本、图片、音频、视频和 PDF 等多模态输入，拥有约 **100 万 Token 上下文窗口**，并可根据任务需求调整不同的思考强度，适合聊天助手、编程开发、文档分析以及大规模 API 调用等场景。\n","icon":"Gemini.Avatar.type={'platform'}","vendor_id":15,"quota_type":1,"model_ratio":0,"model_price":0.035,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini","Gemini混合","Gemini特价"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"claude-opus-4-5-20251101-thinking","description":"claude-opus-4-5-20251101 是 Anthropic 的旗舰 Claude Opus 4.5 的具体版本号，定位为面向真实世界软件工程与高复杂度推理/代理工作流的最强档；官方发布说明它可在 Claude API 里直接用该 ID 调用，并强调其在真实软件工程评测上达到 state-of-the-art，同时给出当时的 API 定价信息以体现“旗舰能力更易用”。","icon":"Claude.Avatar.type={'platform'}","tags":"对话,思考,工具调用","vendor_id":19,"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":["aws","aws限时特价","Claude按量"],"supported_endpoint_types":["anthropic","openai"]}],"group_ratio":{"AWS正价":3.5,"Banana":1,"Banana_GCP":2,"Banana增稳":1.2,"CC MAX":3.2,"Claude Code":1.6,"Claude按量":1,"Gemini":1,"Gemini_flash":1,"Gemini混合":3,"Gemini特价":0.6,"Gemini限时超级特价":0.2,"auto":1,"aws":0.7,"aws限时特价":0.3,"claude code特价":1,"codex":0.5,"test":1,"tol2api":0.6,"ty2api":0.5,"优质GPT":1,"公益渠道":0,"公益渠道2":0,"国内大模型":1,"按次claude":1,"按量Gemini":1,"特价按量Gemini":0.55,"纯血GCP":1.8,"顶级付费Gemini":15},"pricing_version":"a42d372ccf0b5dd13ecf71203521f9d2","success":true,"supported_endpoint":{"0":{"path":"openai","method":"POST"},"STT":{"path":"/v1/audio/transcriptions","method":"POST"},"TTS":{"path":"/v1/audio/speech","method":"POST"},"anthropic":{"path":"/v1/messages","method":"POST"},"audio":{"path":"/v1/audio/transcriptions","method":"POST"},"gemini":{"path":"/v1beta/models/{model}:generateContent","method":"POST"},"image":{"path":"/v1/images/generations/","method":"POST"},"image-generation":{"path":"/v1/images/generations","method":"POST"},"openai":{"path":"/v1/chat/completions","method":"POST"},"openai-response":{"path":"/v1/responses","method":"POST"},"video":{"path":"/v1/videos","method":"POST"},"videos":{"path":"/v1/video/generations","method":"POST"}},"usable_group":{"AWS正价":"aws官方正价，非逆向","Banana":"香蕉绘图分组","Banana_GCP":"GCP按量banana","Banana增稳":"目前最稳banana绘图，空回和不出图均不计费","CC MAX":"纯血Claude MAX号池","Claude Code":"优质CC高可用性，支持1M上下文","Claude按量":"claude默认按量分组，稳定性比aws略高","Gemini":"Gemini默认分组","Gemini_flash":"flash模型默认分组","Gemini混合":"Gemini混合渠道，稳定性优于默认分组","Gemini特价":"Gemini特价渠道，不做稳定性保证","Gemini限时超级特价":"限时超级特价分组，仅支持openai格式，较不稳定","auto":"智能选择","aws":"aws渠道，稳定性低，不推荐","aws限时特价":"限时特价分组，不是很稳定","claude code特价":"claude code特价","codex":"codex渠道，responses端口","test":"测试分组，请勿选择","tol2api":"2api渠道，仅支持openai格式请求，可能有兼容性问题","ty2api":"2api渠道，可能有各种兼容性问题，图像生成有水印","优质GPT":"优质GPT渠道","公益渠道":"公益渠道，限1rpm","公益渠道2":"公益渠道2，限3rpm，不做稳定性保证","国内大模型":"国产大模型专用分组，很不稳定，慎用！","按次claude":"按次计费claude，不保证稳定性和可用性","按量Gemini":"按量gemini分组，稳定性更高，一倍倍率","特价按量Gemini":"特价按量分组，非gcp渠道","纯血GCP":"纯血GCP渠道，1.8倍率","顶级付费Gemini":"付费Gemini渠道"},"vendors":[{"id":31,"name":"Bilibili Index","icon":"Bilibili"},{"id":32,"name":"XiaomiMiMo","icon":"XiaomiMiMo"},{"id":22,"name":"字节跳动","icon":"Doubao.Color"},{"id":2,"name":"Openai","description":"Openai","icon":"OpenAI.Avatar.type={'platform'}"},{"id":15,"name":"Google","icon":"Gemini.Color"},{"id":17,"name":"腾讯","icon":"Hunyuan.Color"},{"id":19,"name":"Anthropic","icon":"Claude.Color"},{"id":21,"name":"酒馆","description":"酒馆聊天专用模型，破限，后缀的数字仅代表不同的线路","icon":"Gemini.Avatar.type={'platform'}"},{"id":29,"name":"Minimax","icon":"Minimax.Avatar.type={'platform'}"},{"id":30,"name":"即梦","icon":"Jimeng.Avatar.type={'platform'}"},{"id":28,"name":"蚂蚁百灵","icon":"AlibabaCloud.Avatar.type={'platform'}"},{"id":24,"name":"OpenAI","icon":"OpenAI"},{"id":5,"name":"DeepSeek","icon":"DeepSeek.Color"},{"id":7,"name":"阿里巴巴","icon":"Qwen.Color"},{"id":20,"name":"xAI","icon":"XAI"},{"id":33,"name":"Mistral","icon":"Mistral.Color"},{"id":34,"name":"[L]客户专属","description":"客户专属供应商，普通用户请勿选择","icon":"Gemini.Color"},{"id":35,"name":"讯飞","icon":"Spark.Color"},{"id":6,"name":"Moonshot","icon":"Moonshot"},{"id":9,"name":"智谱","icon":"Zhipu.Color"},{"id":18,"name":"BAAI","description":"智源研究院","icon":"BAAI.Avatar.type={'platform'}"},{"id":25,"name":"越阶星辰","icon":"Stepfun.Avatar.type={'platform'}"},{"id":26,"name":"Midjourney","icon":"Midjourney.Avatar.type={'platform'}"},{"id":27,"name":"Suno","icon":"Suno.Avatar.type={'platform'}"}]}