{"auto_groups":["default","Codex-Pro2","Codex-Pro3","Grok","Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】"],"data":[{"model_name":"Doubao-1.5-pro-32k","description":"发布日期：2025-01-22\n知识截止：官方未公布\n更新内容：豆包 1.5 代主力模型（32K 上下文版）：稀疏 MoE 架构以较小激活参数实现稠密大模型性能，推理成本低，中文理解与指令遵循表现好。\n适合场景：中文对话与内容创作、客服与营销文案、高并发低成本调用。","vendor_id":7,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":2.5,"cache_ratio":0.2,"create_cache_ratio":0.02125,"enable_groups":["default"],"supported_endpoint_types":["openai"],"pricing_version":"5a90f2b86c08bd983a9a2e6d66c255f4eaef9c4bc934386d2b6ae84ef0ff1f1f"},{"model_name":"kimi-k2.6","description":"发布日期：2026-04-21\n知识截止：官方未公布\n更新内容：通用多模态模型，长程编码任务与 Agent 工作流显著改进；支持思考/非思考双模式，256K 上下文，开源权重。\n适合场景：长任务编码、Agent 工作流、通用多模态对话。","vendor_id":9,"quota_type":0,"model_ratio":3.25,"model_price":0,"owner_by":"","completion_ratio":4.153846153846154,"cache_ratio":0.16923076923076924,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-5-5","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 4 + c * 20 + cr * 0.2 + cc * 5 + cc1h * 8)"},{"model_name":"doubao-seed-2-1-turbo","description":"发布日期：2026-06-23\n知识截止：官方未公布\n更新内容：Seed 2.1 的低成本、低时延版本，功能齐备，效果比肩 Pro 版而价格减半；深度思考模型。\n适合场景：规模化生产：大量线上调用的 Coding、Agent 与多模态理解任务。","vendor_id":7,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.2,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"qwen3.7-max","description":"发布日期：2026-05-20（6 月 8 日快照增加视觉理解）\n知识截止：官方未公布\n更新内容：面向智能体时代的旗舰：编程、办公生产力与长周期自主执行全面领先，SWE-Bench Verified 80.4%；上下文从 256K 提升至 1M。\n适合场景：复杂编码、办公自动化、长周期自主 Agent 工作流。","vendor_id":10,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.1,"create_cache_ratio":1.25,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3-flash","vendor_id":6,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-4-7","description":"发布日期：2026-04-16\n知识截止：2026-01\n更新内容：高难度软件工程显著提升，可放心托管最难的编码任务；长程任务执行更严谨、会自行验证产出后再交付；采用新 tokenizer。\n适合场景：最难的编码问题、长时间自主工程任务、严肃的生产级开发。","vendor_id":5,"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-Max-Temp","KIRO高智商","Claude-AWS","Claude-Max","Claude-Max【不限外接】","CC特别渠道","Claude-AWS-B","Claude-KiroPro","Claude-Kiro【高缓】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"claude-sonnet-4-6","description":"发布日期：2026-02-17\n知识截止：2025-08\n更新内容：编码、计算机操作、长上下文推理、Agent 规划与设计全面升级；SWE-bench Verified 79.6% 接近旗舰水平；1M 上下文（beta），维持 Sonnet 价位。\n适合场景：日常开发、计算机操作 Agent、办公自动化与知识工作。","vendor_id":5,"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":["CC特别渠道","Claude-KiroPro","Claude-Max","Claude-Max-Temp","KIRO高智商","Claude-AWS","Claude-AWS-B","Claude-Kiro【高缓】","Claude-Max【不限外接】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"standard\", p * 3 + c * 15 + cr * 0.3 + cc * 3.75 + cc1h * 6)"},{"model_name":"claude-fable-5-1","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 10 + c * 50 + cr * 0.25 + cc * 12.5 + cc1h * 20)"},{"model_name":"deepseek-v4-pro","description":"发布日期：2026-04-24（预览），2026-08-13 正式版（0813）\n知识截止：2025-05\n更新内容：1.6T 参数（49B 激活）MoE 旗舰，1M 上下文，MIT 开源；0813 正式版 SWE-bench Verified 达 96.4%，Agent 软件工程能力比肩 Opus 4.8；支持非思考/高推理/最大推理三档。\n适合场景：高难度编码与 Agent 工程、超长上下文分析、开源旗舰能力需求。","vendor_id":1,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.008333333333,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:(hour(\"UTC\") \u003e= 1 \u0026\u0026 hour(\"UTC\") \u003c 4 || hour(\"UTC\") \u003e= 6 \u0026\u0026 hour(\"UTC\") \u003c 10) ? tier(\"peak\", p * 9 + cr * 0.30 + c * 27) : tier(\"off_peak\", p * 4.5 + cr * 0.15 + c * 13.5)"},{"model_name":"doubao-seed-2-1-pro","description":"发布日期：2026-06-23\n知识截止：官方未公布\n更新内容：面向 Coding 与 Agent 时代的深度思考旗舰：Coding 工程交付、Agent 长链路执行与多模态理解三大方向全面升级，官方称核心能力比肩 GPT-5.5；更强的需求理解、长期规划与持续修复能力。\n适合场景：复杂编码、长链路 Agent、多步骤工程交付等高复杂度任务。","vendor_id":7,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.2,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.2","description":"发布日期：2026-06-13\n知识截止：官方未公布\n更新内容：上下文从约 200K 跃升至 1M、128K 输出；新增 High/Max 两档思考强度；Code V3 评测全球第三（仅次于 GPT-5.5 与 Opus 4.8）；MIT 开源权重。\n适合场景：超长上下文编码、长程工程交付、开源部署的旗舰替代。","vendor_id":3,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"create_cache_ratio":0.25,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.3-flash","vendor_id":3,"quota_type":0,"model_ratio":0.4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.2875,"enable_groups":["default","特价国模"],"supported_endpoint_types":["openai"]},{"model_name":"kimi-k3","description":"发布日期：2026-07-16（7 月 27 日开放全量权重）\n知识截止：官方未公布\n更新内容：Moonshot 新一代通用旗舰：原生视觉输入、1M 上下文，编码、推理与知识工作全面升级，开源权重。\n适合场景：长文档与多模态理解、编码、研究与知识工作。","vendor_id":9,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"claude-sonnet-5","description":"发布日期：2026-06-30\n知识截止：2026-01\n更新内容：最具 Agent 能力的 Sonnet：可自主规划、调用浏览器与终端，性能接近 Opus 4.8 而价格更低；1M 上下文、128K 输出，默认自适应思考，采用全新 tokenizer。\n适合场景：日常主力模型：编码、Agent 工作流、写作与分析的性价比之选。","vendor_id":5,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】","KIRO高智商","Claude-AWS","Claude-AWS-B","Claude-KiroPro","Claude-Kiro【高缓】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"standard\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5 + cc1h * 4)"},{"model_name":"qwen3.7-flash","description":"发布日期：2026-07-15\n知识截止：官方未公布\n更新内容：Qwen3.7 系列最快、最经济的一档，继承 3.7 代 Agent 与编码能力升级；1M 上下文。\n适合场景：大批量线上调用、轻量 Agent 与工具调用、延迟敏感场景。","vendor_id":10,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:len \u003c= 32000 ? tier(\"input_lte_32k\", p * 0.2 + cr * 0.02 + c * 0.8) : len \u003c= 256000 ? tier(\"input_lte_256k\", p * 0.6 + cr * 0.06 + c * 2.4) : tier(\"input_lte_1m\", p * 1.2 + cr * 0.12 + c * 4.8)"},{"model_name":"qwen3.8-max-0902","vendor_id":10,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.125,"create_cache_ratio":1.25,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-2.5-flash-lite","vendor_id":6,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 0.1 + c * 0.4 + cr * 0.01 + ai * 0.3)"},{"model_name":"gemini-3.1-flash-image","vendor_id":6,"quota_type":1,"model_ratio":0,"model_price":0.16,"owner_by":"","completion_ratio":0,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3.1-flash-lite","vendor_id":6,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-image-2","description":"发布日期：2026-04-21\n知识截止：2025-12\n更新内容：OpenAI 最强图像生成模型（ChatGPT Images 2.0）：生成前先推理，最高 2K 分辨率、单次最多 8 图；文字渲染（含中文）、图像编辑与版式设计能力大幅增强。\n适合场景：海报与设计出图、图文排版、图像编辑与风格化、产品视觉素材。","vendor_id":2,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["Codex-Pro2","Codex-ProFast","生图分组"],"supported_endpoint_types":["image-generation","openai"]},{"model_name":"rmc-6","quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Codex-ProFast"],"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)"},{"model_name":"deepseek-v4.1-flash","vendor_id":1,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default","国模-DeepSeek4.1分组","特价国模"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:(weekday(\"UTC\") \u003e= 1 \u0026\u0026 weekday(\"UTC\") \u003c= 5 \u0026\u0026 ((hour(\"UTC\") \u003e= 1 \u0026\u0026 hour(\"UTC\") \u003c 4) || (hour(\"UTC\") \u003e= 6 \u0026\u0026 hour(\"UTC\") \u003c 10))) ? tier(\"peak\", p * 2 + cr * 0.04 + c * 8) : tier(\"off_peak\", p * 1 + cr * 0.02 + c * 4)"},{"model_name":"qwen3.7-plus","description":"发布日期：2026-06-01\n知识截止：官方未公布\n更新内容：Qwen3.7 多模态 Agent 模型：支持文本/图像/视频输入，可依据视觉参考生成完整交互式网页；成本约为 Max 的三分之一。\n适合场景：多模态 Agent 工作流、视觉驱动的前端开发、图文视频理解。","vendor_id":10,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:len \u003c= 256000 ? tier(\"input_lte_256k_current_discount\", p * 1.6 + cr * 0.16 + cc * 2 + c * 6.4) : tier(\"input_gt_256k_current_discount\", p * 4.8 + cr * 0.48 + cc * 6 + c * 19.2)"},{"model_name":"claude-opus-5","description":"发布日期：2026-07-24\n知识截止：2026-05\n更新内容：以 Fable 5 一半的价格提供接近前沿的智能；Frontier-Bench、GDPval-AA 等编码与知识工作评测新 SOTA，ARC-AGI 3 大幅领先；新增 Fast 模式（约 2.5 倍速）；1M 上下文。\n适合场景：复杂编码与 Agent、企业级工作流、深度分析与文档工作。","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-AWS","Claude-AWS-B","Claude-KiroPro","Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】","KIRO高智商"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"glm-5.3","vendor_id":3,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"create_cache_ratio":0,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"rmc-grok-4.6","vendor_id":8,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Grok"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 200000 ? tier(\"0_200k\", p * 2 + c * 6 + cr * 0.499998) : tier(\"200k_plus\", p * 4 + c * 12 + cr * 1)"},{"model_name":"claude-fable-5","description":"发布日期：2026-06-09（7 月 1 日恢复全面可用）\n知识截止：2026-01\n更新内容：Anthropic 首个公开发布的 Mythos 级模型，定位高于 Opus；几乎所有基准 SOTA（SWE-bench Pro 80.3%），1M 上下文；内置安全分类器，敏感请求自动回退至 Opus 4.8。\n适合场景：最高难度推理与编码、长程 Agent 任务、尖端科研与专业知识工作。","vendor_id":5,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["Claude-Kiro【高缓】","Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】","Claude-AWS","Claude-AWS-B"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 10 + c * 50 + cr * 1 + cc * 12.5 + cc1h * 20)"},{"model_name":"claude-sonnet-5-5","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-Max-Temp","Claude-Max【不限外接】","Claude-Max"],"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":"rmc-opus-5","quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-Max"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"glm-5.2-fast-preview","description":"发布日期：2026-06-23\n知识截止：官方未公布\n更新内容：GLM-5.2 的高速版：同一权重、推理加速优化，输出速度达标准版 1.5～2 倍；1M 上下文、128K 输出，能力对齐标准版。\n适合场景：实时对话、Agent 多轮调用、流式代码生成等延迟敏感场景。","vendor_id":3,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"kimi-k2.5","description":"发布日期：2026-01-27\n知识截止：官方未公布\n更新内容：在推理、编码与 Agent 能力之上新增视觉输入；256K 上下文，开源权重；配合 Agent Swarm 可协调并行子智能体处理更大任务。\n适合场景：多模态理解、编码与 Agent 任务、开源自部署。","vendor_id":9,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5.25,"cache_ratio":0.175,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"grok-4.5","description":"发布日期：2026-07-08\n知识截止：官方未公布\n更新内容：xAI 面向编码、Agent 任务与知识工作的旗舰升级；500K 上下文，支持图像输入、函数调用、联网搜索与 X 搜索、代码执行，推理档位可调。\n适合场景：编码与 Agent 任务、实时信息检索与研究、通用知识工作。","vendor_id":8,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["Grok"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 200000 ? tier(\"0_200k\", p * 2 + c * 6 + cr * 0.3) : tier(\"200k_plus\", p * 4 + c * 12 + cr * 0.6)"},{"model_name":"claude-opus-4-6","description":"发布日期：2026-02-05\n知识截止：2025-08\n更新内容：在 Opus 4.5 基础上提升可靠性与精确度，编码、Agent 与企业工作流表现更稳；新增 Fast 模式（约 2.5 倍速），1M 上下文（beta）。\n适合场景：企业级编码与 Agent、对稳定性要求高的生产任务。","vendor_id":5,"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-Max","Claude-Max-Temp","KIRO高智商","CC特别渠道","Claude-Kiro【高缓】","Claude-Max【不限外接】","Claude-AWS","Claude-AWS-B","Claude-KiroPro"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"claude-sonnet-4-5-20250929","description":"发布日期：2025-09-29\n知识截止：2025-07\n更新内容：发布时最强编码模型（SWE-bench Verified 77.2%），可连续自主工作 30 小时以上；新增上下文编辑与记忆工具，计算机操作能力大幅提升。\n适合场景：长程编码任务、自主 Agent、计算机操作与办公自动化。","vendor_id":5,"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":["KIRO高智商","Claude-AWS","Claude-KiroPro","Claude-Kiro【高缓】","Claude-Max【不限外接】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"standard\", p * 3 + c * 15 + cr * 0.3 + cc * 3.75 + cc1h * 6)"},{"model_name":"grok-4.7","vendor_id":8,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Grok"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 200000\n  ? tier(\"standard\", p * 2 + c * 6 + cr * 0.5)\n  : tier(\"long_context\", p * 4 + c * 12 + cr * 1)"},{"model_name":"claude-haiku-4-5-20251001","description":"发布日期：2025-10-15（快照版本 20251001）\n知识截止：2025-07\n更新内容：小模型达到接近 Sonnet 4 的编码水平，成本约 1/3、速度 2 倍以上；首个支持扩展思考与计算机操作的 Haiku；200K 上下文。\n适合场景：低延迟实时应用、多 Agent 并行子任务、高并发批量处理。","vendor_id":5,"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-Max-Temp","Claude-Max【不限外接】","KIRO高智商","Claude-AWS","Claude-AWS-B","Claude-KiroPro","Claude-Kiro【高缓】","Claude-Max"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"claude-haiku-5-5","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Claude-Max","Claude-Max-Temp","Claude-Max【不限外接】"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 100000\n  ? tier(\"standard\", p * 0.1 + c * 0.5 + cr * 0.01 + cc * 0.125 + cc1h * 0.2)\n  : tier(\"long_context\", p * 0.5 + c * 2.5 + cr * 0.05 + cc * 0.625 + cc1h * 1)"},{"model_name":"gpt-6-sol","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["Codex-Pro2","Codex-Pro3"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5)\n  : tier(\"long_context\", p * 4 + c * 15 + cr * 0.4 + cc * 5)"},{"model_name":"gpt-6.1-sol","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["Codex-Pro2","Codex-Pro3","Codex-ProFast"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 2 + c * 10 + cr * 0.1 + cc * 2.5)\n  : tier(\"long_context\", p * 4 + c * 15 + cr * 0.2 + cc * 5)"},{"model_name":"qwen3.8-max","description":"发布日期：2026-08-03\n知识截止：官方未公布\n更新内容：2.4T 参数 MoE 旗舰：可 10+ 天自主编码交付完整项目；原生多模态视觉贯穿规划—执行—验证全流程；支持 500+ 轮长程闭环优化；1M 上下文。\n适合场景：自主编码与项目交付、法律/金融/设计等专业任务、长程规划。","vendor_id":10,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3-pro-preview","vendor_id":6,"quota_type":0,"model_ratio":0.285,"model_price":0,"owner_by":"","completion_ratio":6.017544,"cache_ratio":0.1,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-4-8","description":"发布日期：2026-05-28\n知识截止：2026-01\n更新内容：修复 4.7 的注释冗长与工具调用问题，默认高思考档位；基准全面提升，工具使用更干净、指令遵循更稳定，适合无人值守运行。\n适合场景：无人值守的自主工程工作流、复杂 Agent、专业知识工作。","vendor_id":5,"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":["KIRO高智商","CC特别渠道","Claude-AWS-B","Claude-Max【不限外接】","Claude-AWS","Claude-KiroPro","Claude-Kiro【高缓】","Claude-Max","Claude-Max-Temp"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"gemini-3.7-flash","vendor_id":6,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 0.75 + c * 3.75 + cr * 0.075)"},{"model_name":"qwen3.6-flash","description":"发布日期：2026-04-16\n知识截止：官方未公布\n更新内容：Qwen3.6 系列高速版：重点提升 agentic coding（多项代码智能体基准大幅超越前代）、数学与代码推理；视觉空间智能、物体定位与目标检测显著增强；1M 上下文。\n适合场景：高并发低成本调用、轻量编码与多模态理解、实时应用。","vendor_id":10,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:len \u003c= 256000 ? tier(\"input_lte_256k\", p * 1.2 + cr * 0.12 + cc * 1.5 + c * 7.2) : tier(\"input_gt_256k\", p * 4.8 + cr * 0.48 + cc * 6 + c * 28.8)"},{"model_name":"gemini-3.1-pro-preview","vendor_id":6,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-4-5-20251101","description":"发布日期：2025-11-24（快照版本 20251101）\n知识截止：2025-08\n更新内容：Opus 系列大幅降价升级（$5/$25），发布时 SWE-bench Verified 最高分；办公文档、表格与计算机操作显著增强，token 效率更高。\n适合场景：复杂编码、企业办公自动化、长程 Agent 任务。","vendor_id":5,"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","Claude-KiroPro","Claude-Kiro【高缓】","Claude-Max【不限外接】","KIRO高智商"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-5.6-luna","description":"发布日期：2026-07-09\n知识截止：2026-02-16\n更新内容：GPT-5.6 家族最快、最省的一档，以极低成本接近 GPT-5.5 峰值表现；1.05M 上下文。\n适合场景：高频低延迟任务：分类打标、路由分流、摘要提取、大批量数据处理。","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-Pro2","Codex-Pro3","Codex-ProFast"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 0.2 + c * 1.2 + cr * 0.02 + cc * 0.25)\n  : tier(\"long_context\", p * 0.4 + c * 1.8 + cr * 0.04 + cc * 0.5)"},{"model_name":"gpt-5.6-terra","description":"发布日期：2026-07-09\n知识截止：2026-02-16\n更新内容：GPT-5.6 家族均衡档，性能与 GPT-5.5 相当而价格约为其一半；1.05M 上下文，支持程序化工具调用与 Multi-agent。\n适合场景：日常开发与办公自动化、规模化生产应用、文档处理与客服等大批量工作负载。","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-Pro2","Codex-Pro3","Codex-ProFast","Codex-Team"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 2.00 + cr * 0.20 + cc * 2.50 + c * 12.00)\n  : tier(\"long_context\", p * 4.00 + cr * 0.40 + cc * 5.00 + c * 18.00)"},{"model_name":"gpt-6-luna","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["Codex-Pro2","Codex-Pro3"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 0.1 + c * 0.5 + cr * 0.01 + cc * 0.125)\n  : tier(\"long_context\", p * 0.2 + c * 0.75 + cr * 0.02 + cc * 0.25)"},{"model_name":"qwen3.8-flash","vendor_id":10,"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"],"supported_endpoint_types":["openai"]},{"model_name":"grok-4.6","description":"发布日期：2026-08-12\n知识截止：2026-02-01\n更新内容：在 4.5 基础上强化长程 Agent 与交互/视觉类工作：DeepSWE 65.9%（前代 54.0%）、Terminal-Bench v3 26.0%；应用与可视化产物的首版生成质量显著提升；500K 上下文。\n适合场景：长时间运行的编码 Agent、跨代码库工程、把想法一次成型为应用或可视化产物。","vendor_id":8,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["Grok"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 200000 ? tier(\"0_200k\", p * 2 + c * 6 + cr * 0.499998) : tier(\"200k_plus\", p * 4 + c * 12 + cr * 1)"},{"model_name":"claude-haiku-4-5","description":"发布日期：2025-10-15\n知识截止：2025-07\n更新内容：小模型达到接近 Sonnet 4 的编码水平，成本约 1/3、速度 2 倍以上；首个支持扩展思考与计算机操作的 Haiku；200K 上下文。\n适合场景：低延迟实时应用、多 Agent 并行子任务、高并发批量处理。","vendor_id":5,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["Claude-Max【不限外接】","Claude-Max","Claude-Max-Temp"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"codex-auto-review","description":"发布日期：随 OpenAI Codex 云端代码审查功能提供\n知识截止：跟随所路由的 Codex 系列模型\n更新内容：面向 Codex 自动代码审查（@codex review / PR 自动审查）的专用通道：审查 PR diff，聚焦 P0/P1 高优先级问题，遵循仓库 AGENTS.md 审查规范。\n适合场景：CI/CD 自动代码审查、PR 质量把关、回归风险排查。","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"cache_ratio":0.1,"enable_groups":["Codex-Pro2.5","Codex-Pro3","Codex-ProFast","Codex-Team","Codex-Pro2"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000\n  ? tier(\"standard\", p * 0.40 + cr * 0.05 + cc * 0.50 + c * 2.40)\n  : tier(\"long_context\", p * 0.80 + cr * 0.08 + cc * 1.00 + c * 3.60)"},{"model_name":"gpt-5.6-sol","description":"发布日期：2026-07-09\n知识截止：2026-02-16\n更新内容：GPT-5.6 家族旗舰，编码、知识工作、网络安全与科学推理全面 SOTA（BrowseComp 92.2%）；新增 ultra 档多智能体并行、程序化工具调用与 Multi-agent；1.05M 上下文，token 效率大幅提升。\n适合场景：最高难度编码与推理、长程 Agent、深度研究、计算机操作与设计任务。","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-ProFast","Codex-Team","Codex-Pro2","Codex-Pro2.5","Codex-Pro3"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 272000 \n  ? tier(\"standard\", p * 5 + cr * 0.5 + cc * 6.25 + c * 30) \n  : tier(\"long_context\", p * 10 + cr * 1.0 + cc * 12.50 + c * 45)"},{"model_name":"gpt-6-astra","vendor_id":2,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["Codex-Team","Codex-Pro2","Codex-Pro2.5","Codex-Pro3","Codex-ProFast"],"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)"},{"model_name":"glm-5.1","description":"发布日期：2026-04-08\n知识截止：官方未公布\n更新内容：744B MoE 开源旗舰（MIT）：首次实现单任务 8 小时、1200+ 步持续自主工作；SWE-Bench Pro 刷新全球最佳，三大硬核代码基准综合全球第三、开源第一；202K 上下文。\n适合场景：长程自主工程任务、复杂系统开发、开源自部署。","vendor_id":3,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:len \u003c 32000 ? tier(\"input_lt_32k\", p * 6 + cr * 1.3 + c * 24) : tier(\"input_gte_32k\", p * 8 + cr * 2 + c * 28)"},{"model_name":"ox-model","quota_type":1,"model_ratio":0,"model_price":0.01,"owner_by":"","completion_ratio":0,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-2.5-flash","vendor_id":6,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333333333333334,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.4","description":"发布日期：2026-03-05\n知识截止：2025-08-31\n更新内容：融合 GPT-5.3-Codex 的顶级编码能力，首个原生支持计算机操作（Computer Use）的通用模型；上下文扩展至 1M tokens，新增工具搜索与可编辑思考计划；推理更省 token、速度更快，幻觉显著降低，表格/演示/文档等办公任务大幅增强。\n适合场景：复杂推理与编码、长程 Agent 任务、专业知识工作（法律/金融/办公文档）、深度研究。","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-ProFast"],"supported_endpoint_types":["openai"]},{"model_name":"kimi-k2.7-code","description":"发布日期：2026-06-12\n知识截止：官方未公布\n更新内容：基于 K2.6 的编码专用分支（1T 总参 / 32B 激活，Modified MIT 开源）：Kimi Code Bench v2 提升 21.8%，推理 token 减少约 30%；仅思考模式，256K 上下文。\n适合场景：编码 Agent、软件工程任务、MCP 工具链工作流。","vendor_id":9,"quota_type":0,"model_ratio":3.25,"model_price":0,"owner_by":"","completion_ratio":4.153846153846,"cache_ratio":0.2,"enable_groups":["default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-2.5-pro","vendor_id":6,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3.6-flash","vendor_id":6,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 0.75 + c * 3.75 + cr * 0.075)"},{"model_name":"gpt-5.3-codex-spark","description":"发布日期：2026-02-12（研究预览）\n知识截止：官方未公布\n更新内容：GPT-5.3-Codex 的小型高速版，OpenAI 首个实时编码模型；运行于 Cerebras WSE-3 硬件，输出超 1000 tokens/s（约为旗舰 15 倍）；128K 上下文，纯文本。\n适合场景：实时结对编程、快速小改动与调试迭代、对延迟极敏感的编码工作流。","vendor_id":2,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["Codex-ProFast","Codex-Pro2.5","Codex-Pro3"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.4-mini","description":"发布日期：2026-03-17\n知识截止：2025-08-31\n更新内容：GPT-5.4 的小型高效版，保留旗舰大部分能力；400K 上下文，支持图像输入、工具调用、网页搜索与计算机操作，速度更快、成本更低。\n适合场景：高并发线上服务、日常对话与轻量编码、成本敏感的批量任务。","vendor_id":2,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["Codex-ProFast","Codex-Team"],"supported_endpoint_types":["openai"]},{"model_name":"deepseek-v4-flash","description":"发布日期：2026-04-24（预览），2026-07-31 正式版（0731）\n知识截止：2025-05\n更新内容：284B 参数（13B 激活）高效 MoE，1M 上下文、最高 384K 输出，MIT 开源权重；以极低成本提供接近旗舰的编码与推理能力。\n适合场景：高性价比日常编码与对话、大规模线上调用、本地/私有化部署。","vendor_id":1,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.02,"enable_groups":["default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"v1:(hour(\"UTC\") \u003e= 1 \u0026\u0026 hour(\"UTC\") \u003c 4 || hour(\"UTC\") \u003e= 6 \u0026\u0026 hour(\"UTC\") \u003c 10) ? tier(\"peak\", p * 3 + cr * 0.10 + c * 9) : tier(\"off_peak\", p * 1.5 + cr * 0.05 + c * 4.5)"},{"model_name":"gemini-3.6-flash-tiered","vendor_id":6,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 200000 ? tier(\"standard\", p * 0.75 + c * 3.75 + cr * 0.075) : tier(\"long_context\", p * 1.5 + c * 7.5 + cr * 0.15)"},{"model_name":"gemini-3.8-flash","vendor_id":6,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["Gemini"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 0.75 + c * 3.75 + cr * 0.075)"},{"model_name":"grok-build-0.1","description":"发布日期：2026-05-14（CLI），05-29 上线 API 公测\n知识截止：官方未公布\n更新内容：xAI 专为 Agent 编码训练的高速模型，驱动 Grok Build CLI；100+ tokens/s、256K 上下文，支持图像输入与工具调用，取代 grok-code-fast-1。\n适合场景：终端编码 Agent、Web 开发与调试、经济型高速工具调用。","vendor_id":8,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["Grok"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c= 200000 ? tier(\"0_200k\", p * 1 + c * 2 + cr * 0.2) : tier(\"200k_plus\", p * 2 + c * 4 + cr * 0.4)"},{"model_name":"gpt-5.5","description":"发布日期：2026-04-23\n知识截止：2025-12-01\n更新内容：新一代旗舰，比 GPT-5.4 更快且更省 token；更懂任务意图，可完整执行“查资料—用工具—自检—交付”的电脑工作闭环；1M 上下文，Codex 中提供 Fast 模式。\n适合场景：复杂编码、专业知识工作、Agent 自动化、长文档分析。","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-Team","Codex-Pro2","Codex-Pro2.5","Codex-Pro3","Codex-ProFast"],"supported_endpoint_types":["openai"]}],"group_ratio":{"Claude-Max":0.8,"Claude-Max-Temp":1.3,"Claude-Max【不限外接】":1.5,"Codex-Pro2":0.15,"Codex-Pro3":0.3,"Codex-ProFast":0.45,"Gemini":0.15,"Grok":0.2,"auto":1,"default":0.18,"特价国模":0.08,"生图分组":1},"pricing_version":"a42d372ccf0b5dd13ecf71203521f9d2","success":true,"supported_endpoint":{"anthropic":{"path":"/v1/messages","method":"POST"},"image-generation":{"path":"/v1/images/generations","method":"POST"},"openai":{"path":"/v1/chat/completions","method":"POST"}},"usable_group":{"Claude-Max":"特价MAX, 随时结束! 禁止蒸馏!!!发现用1罚3","Claude-Max-Temp":"CCMAX【稳定】限CLI 禁止蒸馏","Claude-Max【不限外接】":"CCMAX【稳定】不限外接客户端, 不保证非CLI缓存!禁止蒸馏!!!发现用1罚3","Codex-Pro2":"Pro分组【较稳定】","Codex-Pro3":"Pro分组【稳定】","Codex-ProFast":"Pro分组【比较稳定】, Fast模式ON","Gemini":"Gemini-反重力【稳定】","Grok":"Grok Heavy号池, 智商在线","auto":"自动分组","default":"国模分组, DeepSeek GLM KIMI","特价国模":"特价国模, deepseek-v4.1-flash和glm-5.3-flash","生图分组":""},"vendors":[{"id":10,"name":"阿里巴巴","icon":"Qwen.Color"},{"id":11,"name":"MiniMax","icon":"Minimax.Color"},{"id":2,"name":"OpenAI","icon":"OpenAI"},{"id":3,"name":"智谱","icon":"Zhipu.Color"},{"id":4,"name":"讯飞","icon":"Spark.Color"},{"id":5,"name":"Anthropic","icon":"Claude.Color"},{"id":12,"name":"小米","icon":"Xiaomi"},{"id":13,"name":"快手","icon":"Kling.Color"},{"id":1,"name":"DeepSeek","icon":"DeepSeek.Color"},{"id":6,"name":"Google","icon":"Gemini.Color"},{"id":7,"name":"字节跳动","icon":"Doubao.Color"},{"id":8,"name":"xAI","icon":"XAI"},{"id":9,"name":"Moonshot","icon":"Moonshot"}]}