{"auto_groups":["cheap","default","best"],"data":[{"model_name":"gpt-5.4-mini-2026-03-17","parent_model":"gpt-5.4-mini","description":"gpt-5.4-mini-2026-03-17 是 OpenAI 于 2026-03-19 发布的紧凑型 GPT 模型，支持文本与图像输入、文本输出，知识截止至 2025-08-31。该模型支持推理与工具调用，具备 400000 上下文长度和 128000 最大输出，适合低延迟助手、高并发业务处理以及需要多模态理解的高频工作负载。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1773878400,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"标准\", p * 0.75 + c * 4.5 + cr * 0.075) : tier(\"长上下文\", p * 1.5 + c * 6.75 + cr * 0.15)","pricing_version":"6b91e3c97d19ce094b0b3f7e77d366f5fbf0ad5cda44497e3c7bf95f01ff2020"},{"model_name":"gpt-image-2","description":"gpt-image-2模型。主要应用于营销创意、产品摄影、UI设计、内容创作等领域。\ncheap分组暂不支持4k，2k/4k请切换至best/default分组","icon":"Dalle","tags":"图片,编辑图片","vendor_id":43,"publish_time":1776830615,"is_hot":true,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/21/uptime","quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.25,"image_ratio":1.6,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["image-generation","image-edit"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.04},"default":{"mode":"price","price":0.06}}}},{"model_name":"gpt-image-2-2026-04-21","parent_model":"gpt-image-2","description":"百分百支持4k，仅支持default、best分组，按token收费","tags":"图片,编辑图片","vendor_id":43,"publish_time":1775014054,"enable_creation":true,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.25,"image_ratio":1.6,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["image-generation","image-edit"]},{"model_name":"kimi-k2","description":"Kimi K2 是 Kimi 于 2025-07-11 发布的开源文本模型，支持 256000 上下文长度与最高 16384 输出。它支持工具调用，适合长上下文对话、代码编写以及需要结合外部工具完成任务的智能体场景。","tags":"工具调用,开源","vendor_id":27,"publish_time":1752192000,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["best","default"],"supported_endpoint_types":["openai"]},{"model_name":"nano-banana","description":"Nano Banana 是谷歌推出的图像生成与编辑模型，正式名称为 Gemini 2.5 Flash Image。它支持文生图、基于参考图的自然语言编辑与多图合成，并强调在连续编辑中保持角色和画面一致性，适合创意设计、照片修改和视觉内容迭代。","tags":"视觉,图片","vendor_id":43,"publish_time":1756252800,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-5","description":"Claude Opus 5 是 Claude 推出的旗舰模型，支持文本、图像和 PDF 输入，具备推理与工具调用能力。其上下文长度达 100 万，可用于复杂编程、智能体任务及专业工作，并支持最长 12.8 万文本输出。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1784851200,"enable_creation":true,"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-Bedrock","best","ccmax","cheap","claude","default"],"supported_endpoint_types":["openai","anthropic"]},{"model_name":"gemini-2.5-flash-preview-04-17","description":"Gemini 2.5 Flash Preview 是 Gemini 推出的紧凑型模型，支持文本与图像输入、文本输出及推理，上下文长度为 1,048,756，最大输出 65,536。它侧重低延迟与高吞吐量，适合大规模智能助手和批量处理场景，但不支持工具调用。","tags":"推理,视觉","vendor_id":3,"publish_time":1744848000,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-2.5-pro-exp-03-25","description":"Gemini 2.5 Pro Experimental 0325 是 Gemini 于 2025 年 3 月发布的闭源实验模型，支持文本与图像输入、文本输出，并具备推理能力。模型上下文长度为 1,048,756，最大输出 65,536，适合长文档理解、多模态分析与复杂推理场景，但不支持工具调用。","tags":"推理,视觉","vendor_id":3,"publish_time":1742860800,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-5-convert","parent_model":"gpt-5","description":"gpt5的官方页面逆向出来的api，同等与在官方使用gpt-5","tags":"对话,聊天,文本","vendor_id":43,"publish_time":1774072484,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3.1-flash-image-preview","description":"Nano Banana 2是 Gemini 于 2026 年 2 月发布的图像模型，支持文本、图像和 PDF 输入，并可输出文本与图像，具备推理能力及 65,536 上下文长度。它适合基于提示词的图像生成、图像编辑和视觉设计工作流，但不支持工具调用。","tags":"推理,视觉,图片,编辑图片","vendor_id":3,"publish_time":1772064000,"is_hot":true,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/12/uptime","quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":120,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["openai","gemini"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.4}}}},{"model_name":"imagen-4.0-fast-generate-001","vendor_id":43,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["openai"]},{"model_name":"claude-haiku-4-5","description":"Claude Haiku 4.5 (latest) 是 Claude 于 2025 年 10 月 15 日发布的非开源模型，支持文本、图像和 PDF 输入，输出文本，具备推理与工具调用能力。该模型上下文长度为 200000，最大输出 64000，适合轻量级智能体、办公任务和高响应速度的聊天场景。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1760486400,"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":["cheap"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 1 + c * 5 + cr * 0.1 + cc * 1.25 + cc1h * 2)"},{"model_name":"gemini-3-pro-image-preview-convert-4k","parent_model":"gemini-3-pro-image-preview","description":"Nano Banana Pro的4k版本","tags":"编辑图片,图片","vendor_id":3,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","gemini"],"group_billing":{"groups":{"default":{"mode":"price","price":0.6}}}},{"model_name":"glm-4.5-airx","parent_model":"glm-4.5","description":"GLM-4.5 AirX 是智谱推出的文本模型，面向编码与智能体工程场景，支持工具调用。模型提供 128K 上下文长度和最高 16K 输出，但不支持推理模式，适合长文本处理、代码任务及工具协同应用。","tags":"工具调用","vendor_id":33,"publish_time":1753660800,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.3","description":"智谱 GLM-5.3 是面向长程编码、智能体任务与复杂项目交付的旗舰模型，支持推理和工具调用。模型仅支持文本输入与输出，上下文长度达 1000000，适用于复杂代码开发、需要持续规划的智能体工作流及项目交付场景。","tags":"推理,工具调用","vendor_id":33,"publish_time":1786665600,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"claude-sonnet-4-6","description":"Claude Sonnet 4.6 是 Claude 发布的非开源模型，支持文本、图像与 PDF 输入，并以文本形式输出。该模型具备推理与工具调用能力，适合编码智能体、细致分析以及需要控制生产成本的应用场景。其上下文长度为 1000000，最大输出为 128000。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1771286400,"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-Bedrock","best","ccmax","cheap","claude","default"],"supported_endpoint_types":["openai","anthropic"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 3 + c * 15 + cr * 0.3 + cc * 3.75 + cc1h * 6)"},{"model_name":"gemini-2.5-flash-lite","description":"Gemini 2.5 Flash-Lite 的厂商信息未提供，发布于 2025-06-17，是一款非开源模型，支持文本、图像、音频、视频和 PDF 输入，并输出文本。它支持推理与工具调用，上下文长度为 1048576，最大输出 65536，适合低成本多模态处理、快速智能体和需要长上下文的应用场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1750118400,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.1,"audio_ratio":3,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3-flash-preview","description":"Gemini 3 Flash Preview 的厂商信息未注明，是一款非开源模型，支持文本、图像、音频输入并输出文本，具备推理与工具调用能力。它拥有 1048576 的上下文长度和 65536 的最大输出，适合多模态理解、长上下文分析、自动化工作流与需要较强推理能力的应用场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1765929600,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"audio_ratio":2,"enable_groups":["best","default"],"supported_endpoint_types":["openai","gemini"]},{"model_name":"gemini-3.5-flash-lite","description":"Gemini 3.5 Flash Lite 是 Gemini 推出的快速模型，在成本、速度与多模态推理能力之间取得平衡，支持文本、图像、视频、音频和 PDF 输入，并以文本形式输出。模型支持推理与工具调用，拥有 1,048,576 tokens 上下文窗口，适合长文档分析、多媒体内容理解、智能助手及自动化工作流等场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1784592000,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333333,"cache_ratio":0.1,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"deepseek-v3.1","description":"DeepSeek-V3.1 是 DeepSeek 于 2025-08-21 发布的开源文本模型，支持文本输入与文本输出，知识截止至 2024-07。该模型具备推理与工具调用能力，适合指令遵循、代码生成与分析等场景，并提供 131072 的上下文长度与最大输出长度。","tags":"推理,工具调用,开源","vendor_id":44,"publish_time":1755734400,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"glm-4.5-air","parent_model":"glm-4.5","description":"GLM-4.5-Air 是智谱 GLM 于 2025 年发布的开源文本模型，支持推理与工具调用，上下文长度为 131072，最大输出 98304。作为更轻量的 GLM-4.5 变体，它适合快速代码辅助、自动化智能体和成本敏感的文本处理场景。","tags":"推理,工具调用,开源","vendor_id":33,"publish_time":1753660800,"quota_type":0,"model_ratio":0.3,"model_price":0,"owner_by":"","completion_ratio":6.67,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3-pro","description":"Gemini 3 Pro 是 Gemini 推出的非开源模型，支持文本、图片、视频、音频和 PDF 输入，并以文本形式输出。模型支持推理与工具调用，拥有 1048576 tokens 上下文长度，适合多模态内容理解、复杂分析及需要调用外部工具的任务，也可用于兼容旧版集成。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1763424000,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["best","default"],"supported_endpoint_types":["openai","gemini"]},{"model_name":"gpt-5.6-luna","description":"GPT-5.6 Luna 是 OpenAI 于 2026 年 7 月 9 日发布的非开源模型，支持文本、图像和 PDF 输入，并输出文本。它具备推理与工具调用能力，提供 105 万上下文长度和最高 128000 输出，适合快速、高并发的大批量任务以及需要多模态理解的业务场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1783555200,"enable_creation":true,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000000 ? tier(\"\u003c272k\", p * 0.2 + c * 1.2 + cr * 0.02 + cc * 0.25) : tier(\"272k+\", p * 0.4 + c * 1.8 + cr * 0.04 + cc * 0.5)"},{"model_name":"gemini-3-pro-image-preview-convert","parent_model":"gemini-3-pro-image-preview","description":"Nano Banana Pro 是 Gemini 于 2025-11-20 发布的非开源模型，模型名为 gemini-3-pro-image-preview，支持文本与图像输入，并可输出文本与图像。该模型具备推理能力，面向更高保真度的图像生成与偏设计类的复杂编辑场景，但不支持工具调用。它提供 131072 上下文长度与最高 32768 输出长度，适合长上下文下的多模态创作与视觉设计工作流。\n","vendor_id":3,"quota_type":1,"model_ratio":0,"model_price":0.6,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"group_billing":{"groups":{"default":{"mode":"price","price":0.4}}}},{"model_name":"gpt-4o-search","vendor_id":43,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"qwen3.7-max-2026-05-20","parent_model":"qwen3.7-max","description":"Qwen3.7-Max 是阿里巴巴发布的新一代千问旗舰模型，面向智能体（Agent）场景设计，强调编程、推理和工具调用等能力。它适合用于编程智能体、复杂任务自动执行、办公自动化和企业级工作流等场景。","tags":"推理,工具调用","vendor_id":23,"publish_time":1779235200,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 12 + c * 36 + cr * 2.4)"},{"model_name":"gemini-3-flash","description":"Gemini 3 Flash Preview 是 Gemini 推出的高效多模态推理模型，支持文本、图像、视频、音频和 PDF 输入，并以文本形式输出。模型支持推理与工具调用，拥有 1,048,576 词元上下文窗口和 65,536 词元最大输出，适合长文档分析、多媒体内容理解及复杂任务自动化。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1765929600,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":0.166666,"audio_ratio":2,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"kimi-k2.5","description":"Kimi K2.5最新旗舰模型","tags":"对话,代码,推理","vendor_id":27,"publish_time":1742860800,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":5.25,"cache_ratio":0.175,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"claude-fable-5-1","description":"Claude Fable 5.1 支持文本、图片与 PDF 输入，具备推理和工具调用能力，可处理复杂任务与长流程智能体工作。模型上下文长度达 100 万，适合长文档分析、多模态信息理解、复杂问题推理及需要持续执行的代理场景。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1788220800,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.025,"create_cache_ratio":1.25,"enable_groups":["ccmax","cheap","default"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gemini-2.5-flash-deepsearch-async","vendor_id":3,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["cheap"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"glm-5","description":"GLM-5 是智谱GLM发布的开源旗舰通用模型，支持文本输入与文本输出，具备推理能力和工具调用能力。它拥有 204800 的上下文长度和 131072 的最大输出，适合代码开发、复杂分析以及依赖工具协作的工程类工作流。","tags":"推理,工具调用,开源","vendor_id":22,"publish_time":1770854400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 32000 ? tier(\"0-32k\", p * 4 + c * 18 + cr * 1) : tier(\"32k+\", p * 6 + c * 22 + cr * 1.5)"},{"model_name":"gemini-3-pro-preview","description":"Gemini 3 Pro Preview 是 Gemini 于 2025 年 11 月发布的旗舰预览模型，支持文本、图像、视频、音频和 PDF 多模态输入，并以文本形式输出。模型具备复杂推理、编程和工具调用能力，拥有 1,048,576 Token 上下文窗口，适合长文档分析、多模态内容理解、复杂任务求解与代码开发等场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1763424000,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 200000 ? tier(\"标准\", p * 2 + c * 12 + cr * 0.2) : tier(\"长上下文\", p * 4 + c * 18 + cr * 0.4 + cc * 0.25)"},{"model_name":"kimi-k3","description":"Kimi K3 是 Kimi 推出的开源多模态模型，支持文本、图像和视频输入，并以文本形式输出。模型具备工具调用与可切换的高强度推理能力，拥有约 104 万上下文长度，适合长流程智能体任务、复杂推理及大规模多模态内容分析。","tags":"推理,工具调用,视觉,开源","vendor_id":27,"publish_time":1784160000,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"qwen3.7-max","description":"Qwen3.7 Max 是通义千问于 2026 年 5 月 21 日发布的闭源前沿文本模型，支持最高 1000000 上下文长度与 65536 最大输出。该模型支持推理与工具调用，面向 Agent 框架、编程助手和长任务处理等场景，适合需要长上下文理解、复杂任务规划与文本生成的应用。","tags":"推理,工具调用","vendor_id":23,"publish_time":1779321600,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.2,"create_cache_ratio":1.25,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"deepseek-v4.1-flash","description":"DeepSeek V4.1 Flash 是 DeepSeek 推出的开源多模态模型，支持文本与图像输入、文本输出，具备推理和工具调用能力，专注于复杂推理与智能体编程。模型拥有 100 万 token 上下文窗口，最大输出 38.4 万 token，适合长文档分析、视觉理解、代码生成及自动化智能体等场景。","tags":"推理,工具调用,视觉,开源","vendor_id":44,"publish_time":1788998400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.02,"enable_groups":["cheap","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"gemini-2.5-pro-preview-06-05","description":"gemini-2.5-pro-preview-06-05 是 Gemini 于 2026 年 1 月发布的先进模型，支持文本、图像、视频和音频输入，并以文本形式输出。模型具备复杂推理、编程、多模态分析和工具调用能力，拥有 1,048,576 tokens 上下文长度，适用于长文档处理、代码开发及复杂多模态任务。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"qwen3.5-plus","description":"Qwen3.5 Plus 是通义千问推出的视觉语言模型，支持文本、图像和视频输入，并以文本形式输出。模型具备推理与工具调用能力，拥有 100 万上下文长度和 65536 最大输出，适用于视觉推理、文档理解及智能体任务。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1771200000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 0.8 + c * 4.8 + cr * 0.08 + cc * 1)"},{"model_name":"qwen3.6-flash","description":"Qwen3.6 Flash 是通义千问于 2026-04-27 发布的视觉语言模型，支持文本、图像、视频输入并输出文本，具备推理与工具调用能力。该模型适用于视觉推理、文档理解与智能体任务等场景，支持 1000000 上下文长度与最高 65536 输出长度。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1777248000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"default\", p * 1.2 + c * 7.2)"},{"model_name":"qwen3.7-max-preview","parent_model":"qwen3.7-max","description":"Qwen3.7-Max-Preview 是阿里云旗下千问 Qwen3.7 旗舰模型的预览版，已上线 Qwen Chat 和 Arena AI。搜索结果显示该模型仅支持思考模式，搜索与代码解释器工具暂不可用，适合用于文本问答、复杂推理、数学与代码相关任务的预览体验和评测。","tags":"推理","vendor_id":23,"publish_time":1779148800,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 12 + c * 36 + cr * 2.4)"},{"model_name":"gemini-3-pro-image","description":"Nano Banana Pro 是 Gemini 于 2026 年 5 月 28 日发布的闭源模型，模型名为 gemini-3-pro-image，支持文本与图像输入，并可输出文本与图像。该模型具备推理能力，适合高保真图像生成、设计类重度编辑以及图文结合的创作场景，但不支持工具调用。其上下文长度为 65536，最大输出为 32768，知识截止至 2025 年 1 月。","tags":"推理,视觉,图片,编辑图片","vendor_id":3,"publish_time":1779926400,"is_hot":true,"enable_creation":true,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":60,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai","gemini"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.4}}}},{"model_name":"gemini-3-pro-image-preview-2k","parent_model":"gemini-3-pro-image-preview","description":"逆向于Nano Banana Pro的2k版本","tags":"图片,编辑图片","vendor_id":3,"quota_type":1,"model_ratio":0,"model_price":0.6,"owner_by":"","completion_ratio":0,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"group_billing":{"groups":{"default":{"mode":"price","price":0.6}}}},{"model_name":"gemini-3-pro-image-preview-convert-2k","parent_model":"gemini-3-pro-image-preview","description":"Nano Banana Pro的2k版本","tags":"图片,编辑图片","vendor_id":3,"quota_type":1,"model_ratio":0,"model_price":0.6,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"group_billing":{"groups":{"default":{"mode":"price","price":0.4}}}},{"model_name":"glm-4.6","description":"GLM-4.6 是智谱GLM推出的开源文本模型，支持推理与工具调用，知识截止至 2025 年 4 月。该模型定位于编码智能体、复杂推理和结构化任务等场景，适合用于代码辅助、自动化工作流与文本类任务处理。它支持 204800 tokens 上下文长度，最大输出 131072 tokens，适合处理长文本与大规模生成需求。","tags":"推理,工具调用,开源","vendor_id":33,"publish_time":1759190400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"grok-4.1","description":"grok-4.1 是 xAI 于 2025-11-18 发布的闭源模型，支持文本与图像输入、文本输出，知识截止至 2025-06。该模型支持工具调用，适合代理式工具使用、编码辅助和实时助手等场景，并提供 200000 上下文长度与最高 64000 输出能力。","tags":"工具调用,视觉","vendor_id":47,"publish_time":1763424000,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["cheap"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gemini-3.7-flash","description":"Gemini 3.7 Flash 是 Gemini 推出的高效模型，支持文本、图像、视频、音频和 PDF 输入，并以文本形式输出。它具备推理与工具调用能力，拥有 1,048,576 tokens 的上下文长度，适合智能体工作流、编程及多模态推理场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1786579200,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"audio_ratio":1,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-4.1-2025-04-14","parent_model":"gpt-4.1","description":"GPT-4.1 2025年4月版","tags":"代码,推理","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"o4-mini","description":"o4-mini 是 OpenAI 于 2025 年 4 月 16 日发布的非开源 o 系列模型，支持文本与图像输入、文本输出，具备推理、编码与工具调用能力。该模型定位为快速紧凑的推理模型，适用于代码生成与分析、视觉理解、复杂问题拆解以及需要调用外部工具的智能体场景，支持最高 200000 上下文长度与 100000 最大输出。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1744761600,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai"]},{"model_name":"qwen3.5-flash","description":"Qwen3.5 Flash 是通义千问于 2026-02-23 发布的非开源视觉语言模型，支持文本、图像、视频输入并输出文本，知识截止至 2025-04。该模型具备推理与工具调用能力，适合视觉推理、文档理解以及智能体任务等场景。其上下文长度为 1,000,000，最大输出为 65,536，适合处理长上下文与复杂多模态任务。","icon":"Tongyi","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1771804800,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"o3-mini-2025-01-31","parent_model":"o3-mini","description":"o3-mini 是 OpenAI 于 2025 年 1 月 31 日推出的小型高效推理模型，面向数学、编程和科学等 STEM 场景，强调较低延迟与成本。它支持低、中、高三档推理强度，并提供函数调用、结构化输出等开发者能力，适合在 ChatGPT、API 或智能体应用中处理技术问答、代码与复杂问题求解。","tags":"推理,工具调用","vendor_id":43,"publish_time":1738281600,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-2.5-flash-image","description":"Nano Banana 是 Gemini 推出的图像模型，支持文本与图像输入，并可输出文本和图像，具备推理能力。它擅长快速生成和编辑图像，以及制作角色一致性素材，适用于创意设计、图片修改和角色资产生产等场景；上下文长度与最大输出均为 32768，暂不支持工具调用。","tags":"推理,视觉,图片,编辑图片","vendor_id":3,"publish_time":1756166400,"enable_creation":true,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":100,"cache_ratio":0.25,"audio_ratio":3.333333,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["gemini","image-generation"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.06}}}},{"model_name":"gpt-4.1","description":"GPT-4.1 是 OpenAI 于 2025-04-14 发布的非开源模型，支持文本与图像输入、文本输出，知识截止至 2024-04。该模型支持工具调用，具备约 104 万 tokens 的上下文长度和 32768 tokens 最大输出，适合代码开发、复杂指令遵循以及生产级应用场景。","tags":"工具调用,视觉","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4.1-mini","description":"GPT-4.1 mini 是 OpenAI 于 2025-04-14 发布的非开源模型，支持文本与图像输入、文本输出，并支持工具调用。该模型主打高性价比的快速编码辅助与结构化信息抽取，适合代码生成、文档处理、数据提取和多模态理解等场景。它支持约 104 万上下文长度，最大输出 32768 tokens，知识截止至 2024-04。","tags":"工具调用,视觉","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":0.2,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4.1-mini-2025-04-14","parent_model":"gpt-4.1-mini","description":"OpenAI GPT-4.1 Mini 是 OpenAI 于 2025-04-14 发布的紧凑型 GPT 模型，支持文本与图像输入、文本输出，并支持工具调用。该模型面向低延迟助手和高并发工作负载，适合客服、内容处理、图像理解与自动化流程等场景，提供约 104 万 tokens 的上下文长度和最高 32768 tokens 输出。","tags":"工具调用,视觉","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":0.2,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5","description":"OpenAI GPT-5旗舰模型，编码数学写作全面跃升，幻觉率大幅降低","tags":"对话,代码,推理","vendor_id":43,"publish_time":1723017600,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai-response","anthropic","openai"]},{"model_name":"gpt-5-2025-08-07","parent_model":"gpt-5","description":"GPT-5 2025年8月版","tags":"对话,代码,推理","vendor_id":43,"publish_time":1723017600,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-4.1-nano","description":"GPT-4.1 nano 是 OpenAI 于 2025-04-14 发布的轻量级模型，支持文本与图像输入、文本输出，并支持工具调用。它适合分类、请求路由和超高并发/大批量任务等场景，具备 1,047,576 tokens 上下文长度与最高 32,768 tokens 输出能力，知识截止至 2024-04。","tags":"工具调用,视觉","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.4-nano-2026-03-17","parent_model":"gpt-5.4-nano","description":"gpt-5.4-nano-2026-03-17 是 OpenAI 发布的紧凑型 GPT 模型，支持文本与图像输入、文本输出，具备推理与工具调用能力。该模型提供 400000 上下文长度与 128000 最大输出，适合低延迟助手、高并发和大规模工作负载等场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1773878400,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":6.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"标准\", p * 0.2 + c * 1.25 + cr * 0.02) : tier(\"长上下文\", p * 0.4 + c * 1.875 + cr * 0.04)"},{"model_name":"gpt-image-2.5-sunburst","description":"openai最新gpt-image-2.5模型，更细致，cheap分组不支持2k，4k","vendor_id":43,"publish_time":1788919577,"is_hot":true,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.25,"image_ratio":1.6,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["image-generation","image-edit"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.08},"default":{"mode":"price","price":0.06}}}},{"model_name":"glm-5-turbo","parent_model":"glm-5","description":"GLM-5-Turbo 是智谱GLM于 2026 年 3 月 16 日发布的文本输入、文本输出模型，面向需要更低延迟的编码智能体场景。该模型支持推理与工具调用，适合代码辅助、自动化开发流程、复杂文本任务与智能体应用。模型上下文长度为 200000，最大输出长度为 131072。","tags":"推理,工具调用","vendor_id":22,"publish_time":1773619200,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 32000 ? tier(\"0-32k\", p * 5 + c * 22 + cr * 1.2) : tier(\"32k+\", p * 7 + c * 26 + cr * 1.8)"},{"model_name":"grok-4.2","description":"Grok 4.2最新版","tags":"对话,推理","vendor_id":47,"publish_time":1740758400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["cheap"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"qwen3.5-flash-2026-02-23","parent_model":"qwen3.5-flash","description":"Qwen3.5-Flash API 是阿里巴巴基于 Qwen3.5-35B-A3B 深度优化并在阿里云百炼平台提供的托管生产版本。搜索结果显示其支持长上下文、视觉-语言多模态输入、思考/指令两种推理模式，并内置官方原生工具链，适合对话、快速响应、复杂推理和工具调用类应用。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1771891200,"quota_type":0,"model_ratio":0.6,"model_price":0,"owner_by":"","completion_ratio":10,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"claude-sonnet-5-5","description":"Claude Sonnet 5.5 是 Claude 推出的快速模型，支持文本、图像和 PDF 输入，并可进行推理与工具调用，输出文本。它适合日常编程、智能代理和知识工作等场景，支持 1000000 上下文长度。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1790553600,"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":["claude","default","AWS-Bedrock","best","ccmax"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gemini-3.1-flash-image","description":"Nano Banana 2 是 Gemini 推出的图像模型，支持文本、图像、视频和 PDF 输入，可输出文本与图像，并具备推理能力。它适合提示词驱动的图像生成、图像编辑和视觉设计工作流，上下文长度与最大输出均为 65,536，不支持工具调用。","tags":"推理,视觉,编辑图片,图片","vendor_id":3,"publish_time":1779926400,"is_hot":true,"quota_type":0,"model_ratio":0.25,"model_price":0,"owner_by":"","completion_ratio":120,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","image-generation"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.4}}}},{"model_name":"gpt-4o-mini-2024-07-18","parent_model":"gpt-4o-mini","description":"GPT-4o mini 是 OpenAI 于 2024 年 7 月 18 日发布的轻量级多模态 AI 模型，定位为更高性价比的 GPT-4o 简化版本。搜索结果确认其在 API 中支持文本和视觉输入，具备较强的文本与多模态推理、函数调用能力，适合低成本、高频调用的智能客服、AI Agent、长上下文处理等应用场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1721260800,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-2.5-pro-deepsearch-async","vendor_id":3,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["cheap","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-4","description":"GPT-4基础模型","tags":"对话,代码","vendor_id":43,"publish_time":1678752000,"quota_type":0,"model_ratio":30,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.5,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-6.1-sol","description":"GPT-6.1 Sol 是 OpenAI 推出的模型，支持文本、图像和 PDF 输入，并具备推理与工具调用能力，可输出文本。适用于复杂编程、计算机操作及专业工作等场景，支持约 105 万 token 的上下文长度，最大输出 128000 token。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1790640000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"标准\", p * 2 + c * 10 + cr * 0.1 + cc * 2.5) : tier(\"长上下文\", p * 4 + c * 15 + cr * 0.2 + cc * 5)"},{"model_name":"claude-haiku-4-5-20251001","parent_model":"claude-haiku-4-5","description":"Claude Haiku 4.5 是 Claude 于 2025-10-15 发布的非开源模型，支持文本、图像和 PDF 输入，并输出文本。该模型具备推理与工具调用能力，提供 200000 上下文长度和最高 64000 输出，适合快速响应式助手、分类任务和轻量级智能体等场景。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1760486400,"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","default","AWS-Bedrock","best","ccmax","cheap"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-5.5","description":"GPT-5.5 是 OpenAI 于 2026 年 4 月发布的前沿模型，支持文本、图像和 PDF 输入，输出文本，并具备推理与工具调用能力。该模型适用于编程、计算机使用、研究和知识工作等场景，提供 105 万上下文长度与最高 12.8 万输出长度，适合处理长文档与复杂任务。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1776988800,"is_hot":true,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/12/uptime","quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["azure","best","cheap","default"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"\u003c272k\", p * 5 + c * 30 + cr * 0.5) : tier(\"272k+\", p * 10 + c * 45 + cr * 10)"},{"model_name":"gemini-3.1-flash-image-preview-4k","parent_model":"gemini-3.1-flash-image-preview","description":"逆向于Nano Banana 2  的4k版本","tags":"图片,编辑图片","vendor_id":3,"enable_creation":true,"quota_type":1,"model_ratio":0,"model_price":0.8,"owner_by":"","completion_ratio":0,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"group_billing":{"groups":{"default":{"mode":"price","price":0.5}}}},{"model_name":"gemini-3.1-flash-lite","description":"Nano Banana 2/Gemini 3.1 Flash Lite 是一款未提供厂商信息的闭源 Gemini 模型，发布日期为 2026-05-07，知识截止至 2025-01。它面向低延迟、高吞吐的多模态与智能体工作负载，支持文本、图像、音频输入与文本输出，并具备推理和工具调用能力。该模型适合大规模对话、内容理解、自动化代理等场景，支持 1048576 tokens 上下文和最高 65536 tokens 输出。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1778112000,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"audio_ratio":2,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-4o-2024-11-20","parent_model":"gpt-4o","description":"GPT-4o 2024年11月版本","tags":"对话,多模态","vendor_id":43,"publish_time":1732060800,"kuma_badge_url":"https://status.openai-hub.com/api/badge/14/uptime","quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai-response","anthropic","openai"]},{"model_name":"gpt-5.4-mini","description":"GPT-5.4 Mini 是 OpenAI 于 2026 年 3 月 17 日发布的非开源小型 GPT 模型，支持文本、图像和 PDF 输入，输出文本，并具备推理与工具调用能力。它适合编码子代理、快速调用工具以及高并发/大批量任务处理，提供 400000 上下文长度与最高 128000 输出长度。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1773705600,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["best","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"标准\", p * 0.75 + c * 4.5 + cr * 0.075) : tier(\"长上下文\", p * 1.5 + c * 6.75 + cr * 0.15)"},{"model_name":"qwen3.6-flash-2026-04-16","parent_model":"qwen3.6-flash","description":"qwen3.6-flash 是阿里通义千问团队开发的 Qwen3.6 系列轻量高效模型，对应开源模型 Qwen3.6-35B-A3B，并于 2026 年 4 月发布。搜索结果显示它支持深度推理、智能体编程与原生多模态视觉理解，适用于代码与数学推理、视觉理解、文本生成等场景。","tags":"推理,视觉,开源","vendor_id":23,"publish_time":1776297600,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default","cheap"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 1.2 + c * 7.2)"},{"model_name":"qwen3.6-max-preview","description":"Qwen3.6 Max Preview 是通义千问推出的旗舰文本模型，支持复杂推理、代码任务与工具调用，适合构建智能体工作流。该模型上下文长度为 262144，最大输出 65536，适用于长文档处理、复杂问答、编程辅助和需要外部工具协作的应用场景。","tags":"推理,工具调用","vendor_id":23,"publish_time":1776643200,"quota_type":0,"model_ratio":0.635,"model_price":0,"owner_by":"","completion_ratio":5.992126,"cache_ratio":0.099843,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.1-2025-11-13","parent_model":"gpt-5.1","description":"GPT-5.1 是 OpenAI 推出的 GPT-5 系列增量模型，包含 GPT-5.1 Instant 和 GPT-5.1 Thinking 等形态，强调更自然的对话体验、指令遵循和自适应推理。搜索结果显示其可根据任务复杂度调整思考时间，并支持视频与音频文件等多模态输入，适合日常问答、复杂分析和需要更清晰交互的 ChatGPT 场景。","tags":"推理,视觉,音频","vendor_id":43,"publish_time":1762992000,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["azure","best","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"gpt-5.4","description":"GPT-5.4 是 OpenAI 于 2026 年 3 月 5 日发布的非开源模型，支持文本、图像和 PDF 输入，并输出文本。该模型具备推理与工具调用能力，面向编码、计算机使用等 Agent 工作流，并以较低成本适用于开发辅助、文档理解和自动化任务场景。模型上下文长度为 1,050,000，最大输出为 128,000，知识截止至 2025 年 8 月 31 日。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1741132800,"is_hot":true,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/7/uptime","quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["cheap","default","azure","best"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"\u003c272k\", p * 2.5 + c * 15 + cr * 0.25) : tier(\"272k+\", p * 5 + c * 22.5 + cr * 0.5)"},{"model_name":"gemini-2.5-pro","description":"Gemini 2.5 Pro 是 Google 的推理模型，面向编码、数学与多模态分析场景，支持文本、图像、音频、视频和 PDF 输入，输出为文本。该模型支持工具调用与推理能力，适合复杂问题求解、代码辅助、文档理解和跨媒体内容分析，并提供 1048576 tokens 上下文长度与 65536 tokens 最大输出。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1750118400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 200000 ? tier(\"base\", p * 1.25 + c * 10 + cr * 0.125) : tier(\"第2档\", p * 2.5 + c * 15 + cr * 0.25)"},{"model_name":"o1","description":"o1 是 OpenAI 于 2024 年 12 月 5 日发布的 O 系列推理模型，支持文本和图像输入、文本输出，知识截止至 2023 年 9 月。它面向复杂分析、数学、编程与规划任务，具备推理能力并支持工具调用，适合需要深度思考和步骤化解决问题的场景。该模型上下文长度为 200000，最大输出为 100000。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1733356800,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-4.5","description":"GLM-4.5 是智谱GLM于 2025-07-28 发布的开源文本模型，知识截止至 2025-04，支持推理与工具调用。该模型上下文长度为 131072，最大输出 98304，适合复杂问答、长文档处理、代码与工具协作、需要多步骤推理的文本类应用场景。","tags":"推理,工具调用,开源","vendor_id":33,"publish_time":1753660800,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"qwen3.7-max-2026-05-17","parent_model":"qwen3.7-max","description":"Qwen3.7-Max 是阿里巴巴发布的新一代千问旗舰模型，面向智能体（Agent）场景设计，在编程、复杂推理、多语言理解和工具调用方面有重点提升。搜索结果显示其可执行长周期自主任务并通过 MCP、多智能体协作等方式支持复杂工作流，适用于编程智能体、企业办公自动化和复杂任务执行等场景。","tags":"推理,工具调用","vendor_id":23,"publish_time":1779235200,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 12 + c * 36 + cr * 2.4)"},{"model_name":"qwen3.8-max","description":"Qwen3.8 Max 是通义千问推出的旗舰模型，具备文本、图像、视频和 PDF 多模态理解能力，支持推理与工具调用。模型上下文长度达 100 万，可用于代码开发、专业工作、多模态内容分析及长周期智能体工作流等场景。它仅输出文本，适合需要处理复杂信息并结合外部工具完成任务的应用。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1785715200,"quota_type":0,"model_ratio":6.5,"model_price":0,"owner_by":"","completion_ratio":2.769230769231,"cache_ratio":0.115384615385,"create_cache_ratio":1.153846153846,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-5.6-sol","description":"GPT-5.6 Sol 是 OpenAI 于 2026 年 7 月发布的前沿模型，支持文本、图像与 PDF 输入，并输出文本。它具备推理与工具调用能力，适合复杂专业工作、编程任务和智能体工作流等场景。模型上下文长度达 105 万 tokens，最大输出 12.8 万 tokens，可处理长文档与大规模任务。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1783555200,"enable_creation":true,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["best","cheap","default","azure"],"supported_endpoint_types":["openai-response"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000000 ? tier(\"\u003c272k\", p * 5 + c * 30 + cr * 0.5 + cc * 6.25) : tier(\"272k+\", p * 10 + c * 45 + cr * 1 + cc * 12.5)"},{"model_name":"gemini-3-pro-image-preview","description":"Nano Banana Pro 是 Gemini 推出的图像生成模型，支持文本与图像输入，并可输出文本和图像。模型具备推理能力，擅长高保真图像生成和设计导向的复杂编辑，适用于视觉创作与设计修改场景；上下文长度为 131072，暂不支持工具调用。","tags":"推理,视觉,图片,多模态","vendor_id":3,"publish_time":1763596800,"is_hot":true,"enable_creation":true,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":60,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.4}}}},{"model_name":"gemini-3-pro-image-preview-4k","parent_model":"gemini-3-pro-image-preview","description":"逆向于Nano Banana Pro的4k版本","tags":"图片,编辑图片","vendor_id":3,"enable_creation":true,"quota_type":1,"model_ratio":0,"model_price":1,"owner_by":"","completion_ratio":0,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.8}}}},{"model_name":"gemini-3.5-flash","description":"Gemini 3.5 Flash 是 Gemini 于 2026 年 5 月 19 日发布的非开源模型，支持文本、图像、音频输入并输出文本，兼顾多模态理解、推理能力与工具调用。该模型上下文长度为 1048576，最大输出 65536，适合需要快速处理长上下文、多模态信息分析、自动化工具协作与成本敏感型应用的场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1779160044,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/28/uptime","quota_type":0,"model_ratio":0.75,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"audio_ratio":1,"enable_groups":["best","default"],"supported_endpoint_types":["openai","gemini"]},{"model_name":"kimi-k2.6","description":"Kimi K2.6 是 Kimi 于 2026-04-21 发布的开源多模态模型，支持文本、图像和视频输入，并输出文本。该模型支持推理与工具调用，适合智能体循环、编程任务以及需要理解视觉上下文的应用场景。其上下文长度与最大输出均为 262144，知识截止至 2025-01。","tags":"推理,工具调用,视觉,开源","vendor_id":27,"publish_time":1776729600,"quota_type":0,"model_ratio":0.16,"model_price":0,"owner_by":"","completion_ratio":4.1875,"cache_ratio":0.16875,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"text-embedding-3-small","description":"OpenAI最新小型文本嵌入模型","tags":"嵌入","vendor_id":43,"publish_time":1706140800,"quota_type":0,"model_ratio":0.01,"model_price":0,"owner_by":"","completion_ratio":2,"cache_ratio":0.01,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"deepseek-v4-pro","description":"DeepSeek V4 Pro 是 DeepSeek 于 2026 年 4 月发布的开源文本模型，支持推理与工具调用，具备 100 万 token 上下文长度和最高 384000 token 输出能力。它适合代码生成与分析、长文档处理、多轮复杂任务和长时间智能体运行等场景，知识截止至 2025 年 5 月。","tags":"推理,工具调用,开源","vendor_id":44,"publish_time":1776988800,"quota_type":0,"model_ratio":4.5,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.111111111111,"enable_groups":["ccmax","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"claude-opus-4-5-20251101","parent_model":"claude-opus-4-5","description":"Claude Opus 4.5 是 Claude 于 2025 年 11 月 24 日发布的旗舰模型，支持文本、图像和 PDF 输入，并以文本形式输出。它具备推理与工具调用能力，适合深度推理、编程任务和长周期智能体场景，支持 200000 上下文长度与最高 64000 输出。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1763942400,"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-Bedrock","best","ccmax","cheap","claude","default"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"gpt-5-mini-2025-08-07","parent_model":"gpt-5-mini","description":"GPT-5-mini 是 OpenAI GPT-5 系列中的轻量级模型，搜索结果显示其面向 ChatGPT 与开发者 API 提供，并可作为免费用户超出 GPT-5 配额后的切换模型。它面向成本敏感和中等复杂度任务，支持文本与视觉处理，适合个人开发者、中小企业用于问答、工作流辅助和中等复杂度推理场景。","tags":"推理,视觉","vendor_id":43,"publish_time":1754611200,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"imagen-3.0-fast-generate-001","vendor_id":43,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-4.5-flash","parent_model":"glm-4.5","description":"GLM-4.5-Flash 是智谱推出的开源文本模型，支持推理与工具调用，侧重快速推理、代码生成和智能体工作流。模型拥有 131072 Token 上下文长度，适合长文本处理、编程辅助及需要调用外部工具的自动化任务。","tags":"推理,工具调用,开源","vendor_id":33,"publish_time":1753660800,"quota_type":0,"model_ratio":0.01,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.1","description":"GLM-5.1 是智谱于 2026 年 4 月 14 日发布的开源旗舰模型，支持文本输入与文本输出，具备推理、编码和面向 Agent 工程的能力。该模型支持工具调用，适合复杂问答、代码生成、自动化任务编排与智能体应用等场景，并提供 202752 的上下文长度和最高 128000 的输出长度。","tags":"推理,工具调用,开源","vendor_id":33,"publish_time":1776124800,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["default","cheap"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 32000 ? tier(\"0-32k\", p * 6 + c * 24 + cr * 1.3) : tier(\"32k+\", p * 8 + c * 28 + cr * 2)"},{"model_name":"qwen3.7-plus-2026-05-26","parent_model":"qwen3.7-plus","description":"Qwen3.7-Plus 是阿里巴巴发布的千问 3.7 系列多模态智能体模型，已上线阿里云百炼并提供 API 服务。搜索结果显示，该模型在文本与视觉能力基础上强化了深度推理、自我编程、工具调用、验证测试与自主迭代能力，适用于智能体开发、GUI 自动化、应用复刻、代码生成和多模态理解等场景。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1780358400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 2 + c * 8 + cr * 0.4)"},{"model_name":"o1-2024-12-17","parent_model":"o1","description":"o1 的24年12月17的版本，支持文本和图像输入、文本输出，知识截止至 2023 年 9 月。它面向复杂分析、数学、编程与规划任务，具备推理能力并支持工具调用，适合需要深度思考和步骤化解决问题的场景。该模型上下文长度为 200000，最大输出为 100000。","tags":"推理,代码","vendor_id":43,"publish_time":1734393600,"quota_type":0,"model_ratio":7.5,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"deepseek-v3","description":"DeepSeek V3 是 DeepSeek 于 2024 年 12 月发布的文本输入、文本输出聊天模型，面向指令遵循、代码编写与分析任务。该模型支持工具调用，上下文长度为 65536，最大输出 8192，适用于智能助手、开发辅助、内容分析与自动化工作流等场景。","tags":"工具调用","vendor_id":44,"publish_time":1733011200,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gemini-2.5-flash-deepsearch","vendor_id":3,"quota_type":0,"model_ratio":3,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["cheap"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"ERNIE-4.0-Turbo-8K","description":"ERNIE-4.0-Turbo-8K 是百度自研的文心系列旗舰级大语言模型，已在百度智能云千帆大模型平台新增为模型接口。搜索结果显示其综合效果表现出色，适用于各领域复杂任务场景，并支持自动对接百度搜索以提升问答信息时效性。","tags":"工具调用","vendor_id":24,"publish_time":1719532800,"quota_type":0,"model_ratio":6,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["cheap"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gemini-2.5-pro-deepsearch","vendor_id":3,"quota_type":0,"model_ratio":5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["cheap","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3.1-flash-lite-preview","description":"Nano Banana 2 /Gemini 3.1 Flash Lite Preview 是一款未开源的 Gemini 低延迟预览模型，支持文本、图像、音频、视频和 PDF 输入，并输出文本。它具备推理与工具调用能力，适合高吞吐量的多模态任务和智能体工作负载，可用于内容理解、自动化处理与复杂流程编排等场景。模型上下文长度为 1048576，最大输出长度为 65536。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1772323200,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"audio_ratio":2,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"imagen-4.0-ultra-generate-001","vendor_id":43,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"o1-mini-2024-09-12","parent_model":"o1-mini","description":"o1-mini 是 OpenAI 在 2024 年 9 月发布的 o1 系列轻量级推理模型，定位为更经济高效的小尺寸版本。搜索结果显示它专注于 STEM 推理，尤其适合数学、编码、复杂问题求解等场景，并以更低成本和更快响应满足需要推理但不依赖广泛世界知识的应用。","tags":"推理","vendor_id":43,"publish_time":1726099200,"quota_type":0,"model_ratio":0.605,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"text-embedding-ada-002","description":"text-embedding-ada-002 是 OpenAI 于 2022 年发布的文本嵌入模型，支持文本输入与文本向量表示输出，上下文长度为 8192。该模型适用于语义搜索、检索、聚类与排序等管线，不支持工具调用或推理能力。","vendor_id":43,"publish_time":1671062400,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":0,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["embeddings"]},{"model_name":"gpt-5.4-nano","description":"GPT-5.4 Nano 是 OpenAI 于 2026-03-17 发布的非开源模型，支持文本、图像和 PDF 输入并输出文本，具备推理与工具调用能力。该模型提供 400000 上下文长度和 128000 最大输出，适合简单路由、信息抽取与大批量自动化等成本敏感场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1773705600,"quota_type":0,"model_ratio":0.1,"model_price":0,"owner_by":"","completion_ratio":6.25,"cache_ratio":0.1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"claude-opus-4-7","description":"Opus 4.7在高级软件工程方面相较Opus 4.6有显著提升，尤其在最难的任务上有所提升。用户报告说，能够自信地将最难的编码工作——之前需要密切监督的工作——交接给Opus 4.7。Opus 4.7以严谨和一致性处理复杂且长期运行的任务，精准关注指令，并设计方法在反馈前验证自身输出。\n\n该模型的视野也显著提升：能够以更高分辨率看到图像。在完成专业任务时，它更具品味和创意，能够生成更高质量的界面、幻灯片和文档。而且——虽然它的能力不及我们最强大的模型Claude Mythos Preview——但在多个基准测试中，它比Opus 4.6表现更好：","tags":"编程,对话,推理,多模态","vendor_id":2,"kuma_badge_url":"https://status.openai-hub.com/api/badge/5/uptime","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-Bedrock","best","ccmax","cheap","claude","default"],"supported_endpoint_types":["openai","anthropic"],"billing_mode":"tiered_expr","billing_expr":"tier(\"第一档\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"gpt-5.1","description":"GPT-5.1 是 OpenAI 于 2025-11-14 发布的非开源模型，支持文本、图像、音频输入与输出，并具备推理和工具调用能力。该模型上下文长度为 272000，最大输出为 128000，适用于可控语音生成、旁白、音频交付，以及需要多模态理解与生成的复杂应用场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1763078400,"quota_type":0,"model_ratio":0.625,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["azure","best","cheap","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"gpt-5-nano-2025-08-07","parent_model":"gpt-5-nano","description":"GPT-5-nano 是 OpenAI GPT-5 系列中的轻量级版本，搜索结果显示其随 GPT-5 系列于 2025 年 8 月发布，面向高速度、低延迟和成本敏感的应用场景。结果中提到它支持文本输入输出、图像输入、推理 token、函数调用和结构化输出，适合摘要、分类、移动端或嵌入式等开发需求。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1754524800,"quota_type":0,"model_ratio":0.025,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-6-astra","description":"GPT-6 Astra 是 OpenAI 推出的高能力模型，擅长复杂推理、编程、计算机操作、研究与文档创作。支持文本、图像和 PDF 输入，并支持工具调用与推理，适用于软件开发、资料分析、研究工作及文档处理等场景。模型上下文长度达 1050000，可处理长篇内容与复杂任务。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1788583321,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["openai-response","openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"\u003c272k\", p * 10 + c * 50 + cr * 1 + cc * 12.5) : tier(\"\u003e272k\", p * 20 + c * 75 + cr * 2 + cc * 25)"},{"model_name":"imagen-3.0-generate-002","vendor_id":43,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"o4-mini-2025-04-16","parent_model":"o4-mini","description":"o4-mini 是 OpenAI 发布的轻量级 o 系列推理模型，与 o3 同期推出，定位于在速度、成本和性能之间取得平衡。搜索结果显示它具备多步骤推理、视觉输入理解/视觉推理以及自主调用 ChatGPT 工具的能力，适合编码、数学、科学、视觉理解和高频快速响应场景。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1744761600,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","default"],"supported_endpoint_types":["openai"]},{"model_name":"deepseek-flash","description":"DeepSeek V4.1 Flash 是 DeepSeek 推出的开源多模态模型，支持文本与图像输入、文本输出，具备推理和工具调用能力。该模型面向复杂推理与智能体编程场景，提供 100 万上下文长度，最大输出长度为 393216。","tags":"推理,工具调用,视觉,开源","vendor_id":44,"publish_time":1788998400,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.02,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"text-embedding-3-large","description":"OpenAI最新大型文本嵌入模型","tags":"嵌入","vendor_id":43,"publish_time":1706140800,"quota_type":0,"model_ratio":0.065,"model_price":0,"owner_by":"","completion_ratio":1.428571428569,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"gpt-image-1.5-convert","parent_model":"gpt-image-1.5","description":"逆向于openai的官方模型gpt-image-1.5，等同于在官方页面使用","tags":"图片,逆向,编辑图片","vendor_id":43,"publish_time":1767261146,"quota_type":1,"model_ratio":0,"model_price":0.06,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["image-edit","openai","image-generation"]},{"model_name":"kimi-k2-thinking","parent_model":"kimi-k2","description":"Kimi K2 Thinking 是 Kimi 于 2025-11-06 发布的开源文本模型，支持推理与工具调用，知识截止至 2024-08。该模型面向长周期研究、复杂规划和工具使用等场景，提供 262144 的上下文长度与最大输出长度，适合处理大规模文本任务。","tags":"推理,工具调用,开源","vendor_id":27,"publish_time":1762387200,"quota_type":0,"model_ratio":2,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"claude-opus-4-6","description":"Claude Opus 4.6 是 Claude 于 2026 年 2 月 4 日发布的高端闭源模型，支持文本、图像和 PDF 输入，并输出文本。该模型面向困难编码、复杂规划和较慢但更深入的专家级推理任务，支持工具调用与推理能力。它具备 1000000 上下文长度和 128000 最大输出，适合长文档分析、复杂项目规划、代码辅助与多模态理解等场景。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1751446660,"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":["best","ccmax","cheap","claude","default","AWS-Bedrock"],"supported_endpoint_types":["openai","anthropic"],"billing_mode":"tiered_expr","billing_expr":"tier(\"第一档\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"claude-opus-5-5","description":"Claude Opus 5.5 是一款面向长时间智能体编程与知识工作的非开源模型，支持文本、图像和 PDF 输入，并输出文本。它具备推理与工具调用能力，拥有 100 万上下文长度和最高 12.8 万输出长度，适合处理大型代码库、长文档分析及复杂任务协作。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1790136346,"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-Bedrock","best","ccmax","claude","default"],"supported_endpoint_types":["anthropic","openai"]},{"model_name":"nano-banana-2","description":"Nano Banana 2（Gemini 3.1 Flash Image）是谷歌推出的图像生成与编辑模型，兼顾生成速度、视觉质量和世界知识，并强化了文本渲染、指令遵循及主体一致性。它适用于创意设计、营销素材、图片编辑和内容生产等场景。","tags":"视觉,图片","vendor_id":3,"publish_time":1772150400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3.1-pro","description":"Gemini 3.1 Pro Preview 是 Gemini 于 2026 年 2 月 19 日发布的非开源模型，支持文本、图像、视频、音频和 PDF 输入，并以文本形式输出。模型具备复杂推理、编码、多模态分析和工具调用能力，适用于复杂问题分析、软件开发及多媒体内容理解等场景，支持最长 1,048,576 tokens 的上下文。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1771459200,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"create_cache_ratio":0.1875,"enable_groups":["default","best"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gpt-5-nano","description":"GPT-5 Nano 是 OpenAI 于 2025 年 8 月 7 日发布的紧凑型模型，支持文本与图像输入、文本输出，具备推理能力和工具调用能力。该模型面向低延迟辅助与高并发、大规模工作负载场景，适合客服、内容处理、自动化任务和轻量级多模态应用。其上下文长度为 272000，最大输出为 128000。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1754524800,"quota_type":0,"model_ratio":0.025,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.2","description":"GLM-5.2 是面向长任务时代的旗舰模型。支持真正可用的 1M 上下文，可承载项目级工程上下文，长程任务执行更稳定、工程规范遵循更可靠，一次任务即可完成从需求到多端部署的完整开发链路。","icon":"ChatGLM","tags":"对话,代码","vendor_id":22,"publish_time":1781684603,"enable_creation":true,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":3.5,"cache_ratio":0.25,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response"]},{"model_name":"grok-4","description":"xAI Grok 4 是 xAI 于 2024-07-09 发布的闭源文本模型，支持文本输入与文本输出，知识截止至 2024-07。该模型支持工具调用，具备 256000 上下文长度与最大 256000 输出，适合需要长上下文处理、代码辅助、实时协助和代理式工具使用的场景。","tags":"工具调用","vendor_id":47,"publish_time":1720483200,"quota_type":0,"model_ratio":1.5,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.25,"enable_groups":["cheap"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-5.6-terra","description":"GPT-5.6 Terra 是 OpenAI 于 2026-07-09 发布的非开源模型，支持文本、图像和 PDF 输入，并以文本形式输出。该模型具备推理与工具调用能力，适合日常办公、内容处理、文档理解、信息分析等兼顾能力与成本效率的场景。其上下文长度为 1050000，最大输出为 128000，适合处理长文档和复杂任务。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1783555200,"enable_creation":true,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":8,"enable_groups":["default","azure","best","cheap"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000000 ? tier(\"\u003c272k\", p * 2 + c * 12 + cr * 0.2 + cc * 2.5) : tier(\"272k+\", p * 4 + c * 18 + cr * 0.4 + cc * 5)"},{"model_name":"gpt-4.1-nano-2025-04-14","parent_model":"gpt-4.1-nano","description":"GPT-4.1 nano 是 OpenAI 推出的 GPT-4.1 系列模型之一，通过 API 面向开发者提供，被描述为该系列中速度最快、成本最低的超小型模型。它支持最高 100 万 token 上下文窗口，面向低延迟、高频轻量任务优化，适用于文本分类、自动补全等场景。","vendor_id":43,"publish_time":1744588800,"quota_type":0,"model_ratio":0.05,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4o-mini","description":"GPT-4o mini 是 OpenAI 于 2024-07-18 发布的小型全能模型，支持文本与图像输入、文本输出，并支持工具调用。它具备 128,000 上下文长度和最高 16,384 输出长度，适合低成本多模态辅助、生产级流量处理以及需要调用外部工具的应用场景。","tags":"工具调用,视觉","vendor_id":43,"publish_time":1721260800,"quota_type":0,"model_ratio":0.075,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","default"],"supported_endpoint_types":["openai"]},{"model_name":"o1-preview-2024-09-12","parent_model":"o1-preview","description":"o1-preview-2024-09-12 是 OpenAI 推出的 o1 系列预览版模型，面向 ChatGPT 和 API 用户开放。搜索结果显示，该模型重点强化复杂推理能力，适合数学、科学问题求解以及编程等需要多步骤思考的场景。","tags":"推理","vendor_id":43,"publish_time":1726099200,"quota_type":0,"model_ratio":8.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai-response","anthropic","openai"]},{"model_name":"nano-banana-2-4k","description":"Nano Banana 2 是谷歌推出的图像生成与编辑模型，正式名称为 Gemini 3.1 Flash Image，兼顾较快生成速度、世界知识与语义推理能力。它支持最高 4K 输出、复杂提示词理解、多语言文字呈现及多角色和多对象一致性，适用于创意设计、广告素材、信息图和多元素场景制作。","tags":"推理,视觉,图片","vendor_id":3,"publish_time":1772064000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4o-image-vip","parent_model":"gpt-4o-image","description":"GPT-4o图片VIP高优先级版","tags":"图片,编辑图片","vendor_id":43,"publish_time":1732060800,"quota_type":1,"model_ratio":0,"model_price":0.1,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-image-1-convert","parent_model":"gpt-image-1","description":"逆向于openai的gpt-image-1模型","tags":"图片,编辑图片,逆向","vendor_id":43,"publish_time":1767261095,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["image-edit","openai","image-generation"]},{"model_name":"jev-latest","description":"Jev 是一种面向自动化决策的 AI 模型。它不像 ChatGPT 那样主要生成长篇文本，而是根据输入状态和预设问题，返回结构化结果，例如分类选项、优先级分数、是否执行的概率和置信度。Jev 适合用于客服分流、内容审核、线索评分、Agent 工具调用风控和模型路由等场景，业务程序可以通过 API 直接读取并执行这些决策","tags":"决策","vendor_id":51,"publish_time":1790429653,"is_hot":true,"quota_type":1,"model_ratio":0,"model_price":0.01,"owner_by":"","completion_ratio":0,"enable_groups":["default"],"supported_endpoint_types":["jev"]},{"model_name":"kimi-k3-256k","vendor_id":27,"quota_type":0,"model_ratio":10,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"enable_groups":["best","default"],"supported_endpoint_types":["openai"]},{"model_name":"o3","description":"OpenAI o3最强推理模型，支持深度搜索与工具调用","tags":"推理,代码,数学","vendor_id":43,"publish_time":1744761600,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.25,"enable_groups":["best","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"o3-mini","description":"o3-mini 是 OpenAI 于 2024 年 12 月发布的较小型 o 系列推理模型，支持文本输入与文本输出，具备推理能力和工具调用能力。它适合经济型的代码、数学与规划类任务，可用于需要长上下文处理的应用场景，支持最高 200,000 上下文长度与 100,000 最大输出。","tags":"推理,工具调用","vendor_id":43,"publish_time":1734652800,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"gemini-3.1-flash-lite-image","description":"Nano Banana 2 Lite 是 Gemini 推出的高效多模态图像模型，支持文本与图像输入和输出，主打快速、经济的高并发 1K 图像生成与编辑。模型支持推理和工具调用，拥有 65,536 的上下文长度与最大输出，适合大规模图像生产、编辑及图文交互场景。","tags":"推理,工具调用,视觉,生图","vendor_id":3,"publish_time":1782777600,"quota_type":0,"model_ratio":0.125,"model_price":0,"owner_by":"","completion_ratio":120,"enable_groups":["default","best"],"supported_endpoint_types":["gemini","image-generation"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.06}}}},{"model_name":"gpt-image-2.5-flare","description":"openai最新gpt-image-2.5模型，速度更快，cheap分组不支持2k4k","tags":"图片,编辑图片","vendor_id":43,"publish_time":1788919361,"is_hot":true,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.4,"image_ratio":1.6,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["image-edit","image-generation"],"group_billing":{"groups":{"cheap":{"mode":"price","price":0.08},"default":{"mode":"price","price":0.06}}}},{"model_name":"gpt-5.2-2025-12-11","parent_model":"gpt-5.2","description":"GPT-5.2 是 OpenAI 发布的 GPT-5 系列模型，搜索结果称其于 2025 年 12 月 11 日发布，并提供 Instant、Thinking、Pro 等版本。该模型面向专业知识工作和智能体场景，强调推理、长上下文理解、工具调用、编码与视觉理解能力，适合复杂规划、代码开发、文档/表格/演示生成和多步骤工作流。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1765411200,"quota_type":0,"model_ratio":0.875,"model_price":0,"owner_by":"","completion_ratio":8,"cache_ratio":0.1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-5.4-2026-03-05","parent_model":"gpt-5.4","description":"GPT-5.4 2026年3月版，GPT-5.4 是 OpenAI 于 2026 年 3 月 5 日发布的非开源模型，支持文本、图像和 PDF 输入，并输出文本。该模型具备推理与工具调用能力，面向编码、计算机使用等 Agent 工作流，并以较低成本适用于开发辅助、文档理解和自动化任务场景。模型上下文长度为 1,050,000，最大输出为 128,000，知识截止至 2025 年 8 月 31 日。\n","tags":"对话,代码,推理","vendor_id":43,"publish_time":1741132800,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":6,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"],"billing_mode":"tiered_expr","billing_expr":"len \u003c 272000 ? tier(\"\u003c272k\", p * 2.5 + c * 15 + cr * 0.25) : tier(\"272k+\", p * 5 + c * 22.5 + cr * 0.5)"},{"model_name":"gpt-6-sol","description":"GPT-6 Sol 是 OpenAI 面向复杂编程与智能体工作流的模型，支持文本、图像和 PDF 输入，输出文本，并具备推理与工具调用能力。模型上下文长度达 105 万，适用于复杂代码开发、文档理解及需要调用工具的智能体任务。","tags":"推理,工具调用,视觉","vendor_id":43,"publish_time":1790035200,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":2,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 272000 ? tier(\"标准\", p * 2 + c * 10 + cr * 0.2 + cc * 2.5) : tier(\"长上下文\", p * 4 + c * 15 + cr * 0.4 + cc * 5)"},{"model_name":"claude-fable-5","description":"Claude Fable 5 是 Anthropic 于 2026-06-09 正式发布的最新模型,也是 Anthropic 首款面向公众开放的 Mythos 级模型,能力上限超越此前任何对外发布的 Claude 版本。","tags":"多模态,代码,对话","vendor_id":2,"publish_time":1780940189,"is_hot":true,"enable_creation":true,"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","default","AWS-Bedrock","best","ccmax","cheap"],"supported_endpoint_types":["openai","anthropic"]},{"model_name":"claude-opus-4-8","description":"Claude Opus 4.8 是 Claude 于 2026 年 5 月 28 日发布的高阶模型，支持文本、图像和 PDF 输入，输出文本，并具备推理与工具调用能力。该模型面向高难度推理、编程任务和长周期智能体场景，适合处理复杂分析、代码协作与需要持续规划的工作流。其上下文长度为 1000000，最大输出为 128000，知识截止至 2026 年 1 月。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1779926400,"kuma_badge_url":"https://status.openai-hub.com/api/badge/29/uptime","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-Bedrock","best","ccmax","cheap","claude","default"],"supported_endpoint_types":["openai","anthropic"],"billing_mode":"tiered_expr","billing_expr":"tier(\"默认\", p * 5 + c * 25 + cr * 0.5 + cc * 6.25 + cc1h * 10)"},{"model_name":"claude-sonnet-5","description":"Claude Sonnet 5 是 Claude 于 2026 年 6 月 29 日发布的非开源模型，支持文本、图像和 PDF 输入，并以文本形式输出。该模型具备推理与工具调用能力，适合编码、规划、浏览和日常通用工作等场景。其上下文长度为 100 万，最大输出长度为 128000，知识截止至 2026 年 1 月 31 日。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1782691200,"enable_creation":true,"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":["cheap","claude","default","AWS-Bedrock","best","ccmax"],"supported_endpoint_types":["openai","anthropic"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 3 + c * 15 + cr * 0.3 + cc * 3.75 + cc1h * 6)"},{"model_name":"gemini-3.1-flash-image-preview-2k","parent_model":"gemini-3.1-flash-image-preview","description":"逆向于Nano Banana 2  的2k版本","tags":"图片,编辑图片","vendor_id":3,"quota_type":1,"model_ratio":0,"model_price":0.6,"owner_by":"","completion_ratio":0,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["gemini","openai"],"group_billing":{"groups":{"default":{"mode":"price","price":0.4}}}},{"model_name":"gemini-3.8-flash","description":"Gemini 3.8 Flash 支持文本、图片、视频、音频和 PDF 输入，并具备推理与工具调用能力，可输出文本。模型上下文长度达 1048576，适用于长周期软件工程、自治智能体及复杂企业工作流等场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1788307200,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"audio_ratio":1,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"imagen-4.0-generate-001","vendor_id":43,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai"]},{"model_name":"glm-4.5-x","parent_model":"glm-4.5","description":"GLM-4.5 X 是智谱推出的旗舰文本模型，支持混合推理、代码生成与工具调用，面向智能体工程等复杂任务。模型提供 128K 上下文窗口，最大输出 16384 tokens，适合长文本处理、编程辅助、复杂推理和智能体应用开发。","tags":"推理,工具调用","vendor_id":33,"publish_time":1753660800,"quota_type":0,"model_ratio":4,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"glm-5.3-flash","description":"GLM-5.3-Flash 是智谱推出的开源原生多模态模型，支持文本、图片、视频和 PDF 输入，具备推理与工具调用能力，面向高效编程和长程智能体任务进行了优化。模型支持最长 100 万上下文，适用于代码开发、文档理解、多模态分析以及需要持续规划和执行的智能体场景。","tags":"推理,工具调用,视觉,开源","vendor_id":33,"publish_time":1787702400,"quota_type":0,"model_ratio":0.5,"model_price":0,"owner_by":"","completion_ratio":3,"cache_ratio":0.3,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4o-2024-05-13","parent_model":"gpt-4o","description":"GPT-4o 2024年5月版本","tags":"对话,多模态","vendor_id":43,"publish_time":1715558400,"quota_type":0,"model_ratio":2.5,"model_price":0,"owner_by":"","completion_ratio":3,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"qwen3.7-plus","description":"Qwen3.7 Plus 是通义千问于 2026-06-02 发布的非开源多模态模型，支持文本、图像与视频输入，并生成文本输出。它具备推理与工具调用能力，支持最长 1,000,000 上下文和最高 65,536 输出，适合长上下文智能体、视觉理解、代码相关任务与复杂业务流程处理。","tags":"推理,工具调用,视觉","vendor_id":23,"publish_time":1780358400,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 2 + c * 8 + cr * 0.4)"},{"model_name":"claude-sonnet-4-5-20250929","parent_model":"claude-sonnet-4-5","description":"Claude Sonnet 4.5 是 Claude 于 2025 年 9 月 29 日发布的非开源模型，支持文本、图像与 PDF 输入，输出文本，并具备推理与工具调用能力。该模型定位均衡，适合编程、分析、智能体工作流以及注重成本控制的应用场景，支持最长 1000000 上下文与 64000 最大输出。","tags":"推理,工具调用,视觉","vendor_id":2,"publish_time":1759104000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":5,"enable_groups":["cheap","claude","default","AWS-Bedrock","best","ccmax"],"supported_endpoint_types":["anthropic","openai"],"billing_mode":"tiered_expr","billing_expr":"tier(\"base\", p * 3 + c * 15 + cr * 0.3 + cc * 3.75 + cc1h * 6)"},{"model_name":"gpt-image-2-convert","parent_model":"gpt-image-2","description":"逆向于OpenAI最新最火的图片模型，可使用v1/chat/ path请求","tags":"图片,编辑图片,逆向","vendor_id":43,"publish_time":1776418824,"kuma_badge_url":"https://status.openai-hub.com/api/badge/13/uptime","quota_type":1,"model_ratio":0,"model_price":0.08,"owner_by":"","completion_ratio":0,"enable_groups":["default","best","cheap"],"supported_endpoint_types":["image-generation","image-edit"]},{"model_name":"deepseek-v3.2","description":"DeepSeek-V3.2 是 DeepSeek 于 2025-12-01 发布的开源文本模型，支持 128000 上下文长度与最高 128000 输出，知识截止至 2024-07。该模型支持推理与工具调用，适合指令遵循、代码生成与分析、复杂文本处理等场景。","tags":"推理,工具调用,开源","vendor_id":44,"publish_time":1764547200,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":1.5,"enable_groups":["default","cheap"],"supported_endpoint_types":["openai-response","anthropic","openai"]},{"model_name":"gpt-4o-image","description":"GPT-4o原生图片生成","tags":"图片,编辑图片","vendor_id":43,"publish_time":1732060800,"quota_type":1,"model_ratio":0,"model_price":0.04,"owner_by":"","completion_ratio":0,"enable_groups":["cheap","default"],"supported_endpoint_types":["anthropic","openai","openai-response"]},{"model_name":"grok-4.5","description":"Grok 4.5 是 xAI 推出的新一代模型，支持文本、图像和 PDF 输入，可进行推理与工具调用，并注重降低幻觉风险。它拥有 50 万上下文长度，适用于智能对话、代码开发、文档与图像理解以及智能体工具协作等场景。","tags":"推理,工具调用,视觉","vendor_id":47,"publish_time":1783468800,"is_hot":true,"enable_creation":true,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap","grok"],"supported_endpoint_types":["openai","openai-response"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 200000 ? tier(\"\u003c=200k\", p * 2 + c * 6 + cr * 0.5) : tier(\"200k+\", p * 4 + c * 12 + cr * 1)"},{"model_name":"nano-banana-2-2k","description":"Nano Banana 2 是谷歌发布的图像生成与编辑模型，技术名称为 Gemini 3.1 Flash Image；“2K”表示其默认图像输出规格。模型强调快速生成、复杂指令遵循、文本渲染、主体一致性与世界知识，适合营销素材、社交媒体图片、故事板和创意设计等场景。","tags":"推理,视觉,图片","vendor_id":3,"publish_time":1772064000,"quota_type":0,"model_ratio":37.5,"model_price":0,"owner_by":"","completion_ratio":1,"enable_groups":["cheap"],"supported_endpoint_types":["openai"]},{"model_name":"gpt-4o-2024-08-06","parent_model":"gpt-4o","description":"GPT-4o 2024年8月版本","tags":"对话,多模态","vendor_id":43,"publish_time":1722902400,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gpt-4o","description":"GPT-4o多模态旗舰，支持文本、图像、音频","tags":"对话,多模态","vendor_id":43,"publish_time":1715558400,"quota_type":0,"model_ratio":1.25,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"o1-mini","description":"o1-mini 是 OpenAI 于 2024-09-12 发布的 O 系列推理模型，支持文本输入与文本输出，具备推理和工具调用能力。该模型适合复杂分析、数学、编程与规划等场景，支持 128000 上下文长度，最大输出 65536。","tags":"推理,工具调用","vendor_id":43,"publish_time":1726099200,"quota_type":0,"model_ratio":0.55,"model_price":0,"owner_by":"","completion_ratio":4,"cache_ratio":0.5,"enable_groups":["best","cheap","default"],"supported_endpoint_types":["openai"]},{"model_name":"glm-4.7","description":"GLM-4.7 是智谱GLM发布的开源文本模型，支持推理与工具调用，面向可靠的代码、推理和结构化 Agent 任务。模型上下文长度为 204800，最大输出 131072，适合长文本处理、复杂任务编排、代码辅助与需要工具协同的文本应用场景。","tags":"推理,工具调用,开源","vendor_id":22,"publish_time":1766361600,"quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":4,"enable_groups":["cheap","default"],"supported_endpoint_types":["openai","openai-response","anthropic"]},{"model_name":"gemini-2.5-flash","description":"Gemini 2.5 Flash 是 Gemini 于 2025-06-17 发布的非开源模型，面向对延迟和成本敏感的多模态应用，支持文本、图像、音频、视频和 PDF 输入，并输出文本。该模型支持推理与工具调用，具备 1048576 上下文长度和 65536 最大输出，适合多模态理解、文档处理、智能助理和需要调用外部工具的应用场景。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1750118400,"quota_type":0,"model_ratio":0.15,"model_price":0,"owner_by":"","completion_ratio":8.333333,"cache_ratio":0.1,"create_cache_ratio":1.276667,"audio_ratio":3.333333,"enable_groups":["best","default"],"supported_endpoint_types":["gemini","openai"]},{"model_name":"gemini-3.1-pro-preview","description":"Nano Banana Pro 2 Gemini 3.1 Pro Preview 是 Gemini 发布的非开源预览模型，支持文本、图像、视频、音频与 PDF 输入，并以文本形式输出。该模型强调推理能力，支持工具调用，适合智能体编程、复杂问题求解与多模态信息分析等场景。其上下文长度为 1,048,576，最大输出为 65,536。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1771459200,"is_hot":true,"enable_creation":true,"kuma_badge_url":"https://status.openai-hub.com/api/badge/20/uptime","quota_type":0,"model_ratio":1,"model_price":0,"owner_by":"","completion_ratio":6,"cache_ratio":0.1,"enable_groups":["cheap","default","best"],"supported_endpoint_types":["openai","gemini"],"billing_mode":"tiered_expr","billing_expr":"p \u003c= 200000 ? tier(\"标准\", p * 2 + c * 12 + cr * 0.2) : tier(\"长上下文\", p * 4 + c * 18 + cr * 0.4 + cc * 0.25)"},{"model_name":"gemini-3.6-flash","description":"Gemini 3.6 Flash 是 Gemini 推出的快速模型，支持文本、图像、视频、音频和 PDF 输入，可进行多模态推理与工具调用，并以文本形式输出结果。模型拥有 1,048,576 Token 上下文窗口和最高 65,536 Token 输出，兼顾速度、能力与成本，适合长文档分析、多媒体理解及工具辅助型应用。","tags":"推理,工具调用,视觉,音频","vendor_id":3,"publish_time":1784592000,"quota_type":0,"model_ratio":0.375,"model_price":0,"owner_by":"","completion_ratio":5,"cache_ratio":0.1,"audio_ratio":1,"enable_groups":["best","default"],"supported_endpoint_types":["openai","gemini"]}],"group_description":{"AWS-Bedrock":"Aws-bedrock 企业专供含f5","azure":"部署于Azure的模型，稳定性较高，带有内容审核，适合国内使用","best":"企业专属分组，基于官方 API 中转，稳定性与质量俱佳，为生产环境量身打造。","cccheap":"逆向于claudecode的bug分组，限时开放","ccmax":"ClaudeCode专用分组，高质量，高稳定性，高缓存，适合长时间代码任务","cheap":"涵盖多种低价渠道，质量参差不齐，稳定性有待验证，适合个人用途。","claude":"来自于claude官方key中转，价格较贵且随时变化，加入QQ群获取价格通知","claude-aws":"claude的aws分组，限时开放，有货则开放","claude-kiro":"逆向于kiro的claude模型，包含最新的claude，成本最低，可能降智","codex":"限时开放，价格便宜，逆向于Codex，可能降智不稳定","default":"很少出问题，多种优质渠道负载均衡，性能稳定可靠，质量优异，基本不降智，适合高频使用。","grok":"grok专用分组，可用grok模型"},"group_ratio":{"AWS-Bedrock":3.5,"auto":1,"azure":2,"best":6.8,"cccheap":999,"ccmax":2,"cheap":1,"claude":4.5,"claude-aws":1,"claude-kiro":0.35,"codex":0.3,"default":3,"grok":0.4},"pricing_version":"a42d372ccf0b5dd13ecf71203521f9d2","search_keywords":["gpt-image","claude-opus-5","gpt-5.6","gpt-5.5","nano banana","nano banana pro","nano banana 2","claude-fable-5","glm-5.2","图片","编程","对话"],"success":true,"supported_endpoint":{"anthropic":{"path":"/v1/messages","method":"POST","data":{"max_tokens":1024,"messages":[{"content":"Hello!","role":"user"}],"model":"{model}"},"data_type":"application/json"},"audio-transcription":{"path":"/v1/audio/transcriptions","method":"POST","data":{"file":"@audio.mp3","model":"{model}"},"data_type":"multipart/form-data"},"embeddings":{"path":"/v1/embeddings","method":"POST","data":{"input":"The quick brown fox","model":"{model}"},"data_type":"application/json"},"gemini":{"path":"/v1beta/models/{model}:generateContent","method":"POST","data":{"contents":[{"parts":[{"text":"Hello!"}]}]},"data_type":"application/json"},"image-edit":{"path":"/v1/images/edits","method":"POST","data":{"image":"@image.png","mask":"@mask.png","model":"{model}","prompt":"Add a hat to the cat"},"data_type":"multipart/form-data"},"image-generation":{"path":"/v1/images/generations","method":"POST","data":{"model":"{model}","n":1,"prompt":"A cute baby sea otter","size":"1024x1024"},"data_type":"application/json"},"jev":{"path":"/v1/chat/completions","method":"POST","data":{"model":"{model}","questions":{"churn":{"criteria":{"false":"只是想把问题解决。","true":"提到取消或换用别家。"},"instructions":"客户是否威胁要取消服务或转投竞争对手？","type":"noul"},"frustration":{"criteria":["平静","不满但礼貌","非常不满"],"instructions":"客户听起来有多不满？","type":"score"},"team":{"criteria":{"技术":"报错或产品功能无法正常工作。","账单":"发票、付款、订阅或退款。","账户":"登录、访问权限或账户设置。"},"instructions":"这张工单应该由哪个团队处理？","type":"choice"},"today":{"criteria":{"false":"可以排进正常队列。","true":"有今天的截止日期或业务影响。"},"instructions":"这条消息描述的问题是否需要今天处理？","type":"noul"}},"state":"从昨天开始，我们控制台的每次 CSV 导出都中途失败。我已经重试了三次，还是拿不到完整报表。今天下午我们要做客户评审。能不能帮忙在评审前把导出修好？"},"data_type":"application/json"},"openai":{"path":"/v1/chat/completions","method":"POST","data":{"messages":[{"content":"Hello!","role":"user"}],"model":"{model}"},"data_type":"application/json"},"openai-realtime":{"path":"/v1/realtime?model={model}","method":"WSS","data_type":"application/json"},"openai-response":{"path":"/v1/responses","method":"POST","data":{"input":"Hello!","model":"{model}"},"data_type":"application/json"},"openai-video":{"path":"/v1/videos","method":"POST","data":{"model":"{model}","prompt":"A cat playing piano"},"data_type":"application/json"}},"tool_prices":{"file_search":2.5,"google_search":14,"web_search":10,"web_search:gpt-4.1*":25,"web_search:gpt-4o*":25,"web_search_20250305":10,"web_search_baidu":25,"web_search_preview":10,"web_search_preview:gpt-4.1*":25,"web_search_preview:gpt-4.1-mini*":25,"web_search_preview:gpt-4o*":25,"web_search_preview:gpt-4o-mini*":25},"usable_group":{"AWS-Bedrock":"Aws-bedrock","auto":"自动选择","azure":"Azure的模型","best":"企业分组","cccheap":"cc专用【限时分组】","ccmax":"ClaudeCode专用","cheap":"便宜分组","claude":"claude官方","claude-aws":"AWS的Claude","claude-kiro":"kiro逆向【限时开放】","codex":"Codex逆向【限时分组】","default":"默认分组","grok":"Grok专用"},"vendors":[{"id":1,"name":"OpenAI","description":"OpenAI官方API模型","icon":"OpenAI"},{"id":6,"name":"Midjourney","icon":"Midjourney"},{"id":35,"name":"阿里巴巴","icon":"Qwen.Color"},{"id":39,"name":"百度","icon":"Wenxin.Color"},{"id":5,"name":"Suno","description":"AI生成音乐","icon":"Suno"},{"id":31,"name":"Pika","description":"Pika AI视频生成","icon":"Pika"},{"id":36,"name":"Moonshot","icon":"Moonshot"},{"id":37,"name":"讯飞","icon":"Spark.Color"},{"id":38,"name":"MiniMax"},{"id":29,"name":"Runway","description":"Runway AI视频生成和编辑","icon":"Runway"},{"id":3,"name":"Gemini","description":"谷歌公司的AI模型","icon":"Gemini"},{"id":22,"name":"智谱GLM","description":"智谱AI大模型，GLM系列","icon":"Zhipu"},{"id":25,"name":"讯飞星火","description":"科大讯飞星火大模型","icon":"iFLYTEK"},{"id":26,"name":"Flux","description":"Black Forest Labs出品的AI图片生成模型","icon":"Flux"},{"id":30,"name":"可灵","description":"快手可灵AI，图片和视频生成","icon":"Kling"},{"id":43,"name":"OpenAI","description":"OpenAI官方API模型","icon":"OpenAI"},{"id":44,"name":"DeepSeek","icon":"DeepSeek.Color"},{"id":7,"name":"xAI","description":"一家致力于探索宇宙本质并开发强大AI模型（如Grok）的人工智能公司","icon":"XAI"},{"id":48,"name":"快手","icon":"Kling.Color"},{"id":51,"name":"Jev"},{"id":23,"name":"通义千问","description":"阿里云通义千问系列模型","icon":"Qwen"},{"id":27,"name":"Kimi","description":"月之暗面Kimi系列模型","icon":"Moonshot"},{"id":32,"name":"Anthropic","icon":"Claude.Color"},{"id":34,"name":"Google","icon":"Gemini.Color"},{"id":47,"name":"xAI","description":"一家致力于探索宇宙本质并开发强大AI模型（如Grok）的人工智能公司","icon":"XAI"},{"id":2,"name":"Claude","description":"Claude模型，适合代码","icon":"Claude"},{"id":4,"name":"DeepSeek","icon":"DeepSeek"},{"id":41,"name":"Anthropic","icon":"Claude.Color"},{"id":42,"name":"Google","icon":"Gemini.Color"},{"id":46,"name":"讯飞","icon":"Spark.Color"},{"id":45,"name":"百度","icon":"Wenxin.Color"},{"id":49,"name":"阿里巴巴","icon":"Qwen.Color"},{"id":50,"name":"零一万物","icon":"Yi.Color"},{"id":24,"name":"百度文心","description":"百度文心一言ERNIE系列","icon":"Baidu"},{"id":33,"name":"智谱","icon":"Zhipu.Color"},{"id":40,"name":"快手","icon":"Kling.Color"}]}