{"data":{"HeaderNavModules":"","SidebarModulesAdmin":"","announcements":[{"content":"最新上架 Kiro 分组（0.3 倍率），Kiro 正价满血号池，全功能稳定、高智商，CC缓存命中率 90+","extra":"","id":49,"publishDate":"2026-11-11T02:35:28.135Z","type":"success"},{"content":"最新上线 claude-fable-5-2，Claude 相关各分组均可调用","extra":"","id":48,"publishDate":"2026-11-10T12:02:53.000Z","type":"success"},{"content":"Image-2.6 绘画模型已上线 gpt-image-2.6-sunburst、gpt-image-2.6-flare Mini分组特价 0.2元/张","extra":"","id":47,"publishDate":"2026-11-05T08:15:32.105Z","type":"success"},{"content":"🚀 国产高性价比分组上线：DeepSeek 0.15倍率 自建开源模型号池资源","extra":"","id":46,"publishDate":"2026-11-05T07:21:03.958Z","type":"success"},{"content":"CC/Codex 一键配置工具已上线 https://wolfai.apifox.cn/9603230m0","extra":"","id":45,"publishDate":"2026-10-28T07:43:29.774Z","type":"success"},{"content":"最新支持 gpt-6.1-terra 模型，已上线可直接调用","extra":"","id":44,"publishDate":"2026-10-27T09:57:38.000Z","type":"success"},{"content":"Codex 分组升级 OpenAI官Key 90+缓存 稳定高速不降智","extra":"","id":43,"publishDate":"2026-10-16T10:51:17.152Z","type":"success"},{"content":"新增DeepSeek分组，0.15倍率 DeepSeek/Kimi//MiniMax/GLM/Qwen 国产开源模型","extra":"","id":42,"publishDate":"2026-10-16T10:49:56.000Z","type":"success"},{"content":"gemini-4-argon 已经支持","extra":"","id":41,"publishDate":"2026-10-14T10:44:29.768Z","type":"success"},{"content":"最新模型 gpt-6.1-sol 已支持","extra":"","id":40,"publishDate":"2026-09-30T17:39:29.548Z","type":"default"},{"content":"最新模型 claude-sonnet-5-5 已支持","extra":"","id":39,"publishDate":"2026-09-29T10:50:50.169Z","type":"success"},{"content":"新模型 gpt-6-sol 、 gpt-6-luna 、claude-opus-5-5、grok-4.7 均已支持","extra":"","id":38,"publishDate":"2026-09-23T02:41:39.160Z","type":"success"},{"content":"新增GPT Imge-2.5模型，快速版：gpt-image-2.5-flare，高质量：gpt-image-2.5-sunburst","extra":"","id":37,"publishDate":"2026-09-09T08:26:14.466Z","type":"success"},{"content":"claude-fable-5-1、gemini-3.8-flash 已经上线！","extra":"","id":36,"publishDate":"2026-09-03T11:26:23.517Z","type":"success"},{"content":"Codex 升级至 0.149.0 后，如遇 API_KEY_REQUIRED 或 401 Unauthorized，请打开 config.toml 配置文件，加入这段配置：requires_openai_auth = true，保存并重启 Codex 即可恢复正常使用。","extra":"","id":35,"publishDate":"2026-08-21T09:09:59.973Z","type":"warning"},{"content":"gemini-3.7-flash、deepseek-v4-pro-0813 已经支持","extra":"","id":34,"publishDate":"2026-08-14T06:55:31.668Z","type":"success"},{"content":"grok-4.6 已支持，推荐使用 Codex 分组和客户端接入使用","extra":"","id":33,"publishDate":"2026-08-14T06:55:12.839Z","type":"success"},{"content":"微信登陆方式暂时恢复，请尽快在个人资料页绑定邮箱，然后在登陆页重置密码后用邮箱登陆，微信登陆方式即将彻底下线！","extra":"","id":32,"publishDate":"2026-08-12T06:24:08.080Z","type":"warning"},{"content":"claude-opus-5 已支持，Claude Code 最新版可用 /model 命令切换","extra":"","id":30,"publishDate":"2026-07-26T03:58:16.013Z","type":"success"},{"content":"最新支持 Gemini 3.6 Flash、Gemini 3.5 Flash-Lite ","extra":"","id":29,"publishDate":"2026-07-22T03:59:55.600Z","type":"success"},{"content":"GPT-5.6 Sol、Terra、Luna已支持，Codex 最新版可选择调用","extra":"","id":28,"publishDate":"2026-07-10T01:53:07.423Z","type":"success"},{"content":"claude-fable-5 全面解禁上线，claude-sonnet-5 已支持","extra":"","id":27,"publishDate":"2026-07-02T06:51:12.860Z","type":"success"},{"content":"claude-fable-5 已支持，默认和Code分组可用","extra":"","id":26,"publishDate":"2026-06-10T03:50:17.127Z","type":"success"},{"content":"Codex 官方已下线 GPT-5.2 和 GPT-5.3-Codex ，仅支持GPT-5.4 和 GPT-5.5 接入使用","extra":"","id":25,"publishDate":"2026-06-02T01:57:08.740Z","type":"warning"},{"content":"claude-opus-4-8 已支持，Claude Code 最新版已支持接入，切换命令：/model","extra":"","id":24,"publishDate":"2026-05-29T02:12:26.713Z","type":"success"},{"content":"Code分组倍率下调为0.9，Codex分组倍率下调为0.25，Azure 分组倍率调整为0.35","extra":"","id":23,"publishDate":"2026-05-25T11:58:17.445Z","type":"default"},{"content":"最新支持 gemini-3.5-flash，Gemini 分组可用","extra":"","id":22,"publishDate":"2026-05-20T11:33:00.083Z","type":"success"},{"content":"默认和Azure分组最新支持 deepseek-v4-flash、deepseek-v4-pro","extra":"","id":21,"publishDate":"2026-04-26T13:52:41.391Z","type":"success"},{"content":"Gemini 分组价格上调，Vip、Svip分组上线Gemini优质渠道","extra":"","id":20,"publishDate":"2026-04-26T13:49:27.130Z","type":"default"},{"content":"gpt-5.5 已支持，Codex可在配置文件中修改model或使用“codex -m gpt-5.5”启动","extra":"","id":19,"publishDate":"2026-04-24T01:43:03.318Z","type":"success"},{"content":"最新支持 OpenAI 绘画模型 gpt-image-2","extra":"","id":18,"publishDate":"2026-04-22T07:42:40.560Z","type":"success"},{"content":"claude-opus-4-7 模型上线，CC 升级最到新版即可使用","extra":"","id":17,"publishDate":"2026-04-17T05:09:19.103Z","type":"success"},{"content":"最新上线“Codex”分组，Codex/Claude Code 特价逆向渠道，不限客户端使用","extra":"","id":16,"publishDate":"2026-03-07T03:16:35.074Z","type":"success"},{"content":"最新上线“Gemini”分组，Gemini CLI/Google AI Studio/Web 逆向混合渠道","extra":"","id":15,"publishDate":"2026-03-07T03:15:06.813Z","type":"success"},{"content":"最新支持 gpt-5.4，Codex 可在配置文件中修改model或使用“codex -m gpt-5.4”启动","extra":"","id":14,"publishDate":"2026-03-06T01:33:54.204Z","type":"success"},{"content":"最新上线\"Code“分组，官方Claude Max号池，仅支持Claude Code客户端调用","extra":"","id":13,"publishDate":"2026-03-05T05:56:40.627Z","type":"success"},{"content":"最新支持 gemini-3.1-pro-preview","extra":"","id":12,"publishDate":"2026-02-21T15:23:23.229Z","type":"success"},{"content":"最新支持 claude-sonnet-4-6","extra":"","id":11,"publishDate":"2026-02-19T03:46:57.037Z","type":"success"},{"content":"最新上线 claude-opus-4-6，claude-opus-4-6-20260205，Codex 已支持 gpt-5.3-codex","extra":"","id":10,"publishDate":"2026-02-06T02:23:25.707Z","type":"success"},{"content":"最新支持 GPT-5.2 系列模型：gpt-5.2-codex，gpt-5.2-xhigh，gpt-5.2-pro；","extra":"","id":9,"publishDate":"2025-12-20T02:44:11.520Z","type":"success"},{"content":"最新支持 gemini-3-flash-preview","extra":"","id":8,"publishDate":"2025-12-18T08:58:10.813Z","type":"success"},{"content":"最新支持 claude-opus-4-5-20251101，按次计费 cs-claude-opus-4-5","extra":"","id":7,"publishDate":"2025-11-25T02:06:17.174Z","type":"success"},{"content":"支持gemini-3-pro-image，谷歌Nano Banana Pro最先进图像生成与编辑模型","extra":"","id":6,"publishDate":"2025-11-21T04:58:59.411Z","type":"success"},{"content":"Gemini3 模型上线，gemini-3-pro-preview ","extra":"","id":5,"publishDate":"2025-11-19T01:33:23.190Z","type":"success"},{"content":"最新支持 gpt-5.1-codex、gpt-5.1 模型，Codex 支持调用","extra":"","id":4,"publishDate":"2025-11-14T01:31:52.000Z","type":"success"},{"content":"上线 claude-haiku-4-5 系列模型，速度更快价格更低，性能与Sonnet 4相当","extra":"","id":1,"publishDate":"2025-10-16T07:14:37.000Z","type":"success"},{"content":"支持 sora_video2 视频模型，Chat 聊天格式即可调用","extra":"","id":2,"publishDate":"2025-10-01T07:15:35.000Z","type":"success"},{"content":"最新支持 claude-sonnet-4-5-20250929，按次计费模型 cs-claude-sonnet-4-5","extra":"","id":3,"publishDate":"2025-09-30T07:15:49.000Z","type":"success"}],"announcements_enabled":true,"api_info":[{"color":"red","description":"国内高速直连","id":1,"route":"HK香港线路","url":"https://wolfai.top"},{"color":"blue","description":"海外加速负载优化","id":2,"route":"海外CDN线路","url":"https://api.hdgsb.com"},{"color":"green","description":"全球加速负载优化","id":3,"route":"全球CDN线路","url":"https://www.wolfai.top"}],"api_info_enabled":true,"chats":[{"Image2-Video":"https://d.hdgsb.com/#/?settings={\"key\":\"{key}\",\"url\":\"{address}\"}"},{"Midjourney-Image":"https://vercel.ddaiai.com/#/?settings={\"key\":\"{key}\",\"url\":\"{address}\"}"}],"checkin_enabled":false,"custom_currency_exchange_rate":1,"custom_currency_symbol":"¤","data_export_default_time":"hour","default_collapse_sidebar":false,"default_use_auto_group":false,"demo_site_enabled":false,"discord_client_id":"","discord_oauth":false,"display_in_currency":true,"docs_link":"https://wolfai.apifox.cn","email_verification":true,"enable_batch_update":true,"enable_data_export":true,"enable_drawing":true,"enable_task":true,"faq":[{"answer":"这是正常现象。很多 Agent 客户端并不是“一条用户消息 = 一次模型请求”。\n\n例如 Claude CLI、Codex Desktop 这类 Agent 工具，在完成一个任务时，可能会连续发起多轮模型请求：\n\n- 理解需求\n- 读取上下文\n- 调用工具\n- 总结结果\n- 继续规划下一步\n\n此外，有些客户端还会在你没有明确发送新消息时，自动发起一些辅助请求，例如：\n\n- 生成首页推荐内容\n- 加载或总结上下文\n- 检查当前会话状态\n- 为下一步操作做预处理\n\n账单明细展示的是**底层真实模型请求**，所以你可能会看到同一时间附近出现多条记录。\n\n### 简单理解\n\n\u003e **你看到的是一次操作或一次对话，系统记录的是后台所有真实发生的模型调用。**","id":1,"question":"为什么我只发了一条消息，账单里却有多条记录？"},{"answer":"Agent 工具可能会把一个任务拆成多个步骤，不同步骤使用不同模型。例如一个步骤用更强模型做规划，另一个步骤用更轻量模型处理简单内容。\n\n所以同一轮任务中出现多个模型，不一定代表异常，而是 Agent 客户端在按步骤调用不同模型。","id":2,"question":"为什么模型选择 gpt-5.5，但是账单中出现了 gpt-5.4-mini？"},{"answer":"Cache 是模型服务的上下文缓存机制。Agent 在连续多轮任务中，经常会重复携带大量上下文，例如项目文件、历史对话、工具结果等。\n\n- Cache 写入：首次把上下文写入缓存，通常费用较高。\n- Cache 读取：后续复用已缓存上下文，通常比重新输入便宜很多。\n\n所以你可能会看到某些记录的 Cache 写入很多、费用较高；另一些记录 Cache 读取很多、费用较低。这通常说明 Agent 在复用上下文缓存。","id":3,"question":"Cache 读取和 Cache 写入是什么？为什么金额差异很大？"},{"answer":"因为模型计费不只看普通输入和输出 Tokens，还可能包含 Cache 写入、Cache 读取、模型单价差异等因素。\n尤其是 Claude / Agent 场景中，大量费用可能来自上下文缓存，而不是你本次手动输入的那几句话。","id":4,"question":"为什么有些请求输入 Tokens 很少，但费用不低？"},{"answer":"“应扣金额”可以理解为按模型官方或标准价格计算的参考金额；“实扣金额”是按当前使用分组折扣后实际从余额中扣除的金额。\n\n如果页面显示 “2.8折”“6折”“1.1折” 等标签，表示本次请求享受了对应折扣。","id":5,"question":"为什么应扣金额和实扣金额不一样？"},{"answer":"- 在同一个会话里持续对话，客户端通常会把更多历史消息、文件内容、工具结果等一起带给模型，单次请求的上下文会变长，因此体感上每次消耗会越来越高。建议在长对话后适当开启新会话、压缩上下文；如果只是简单问题，可以选择更轻量的模型，以降低单次消耗。\n- 如果在一个会话里频繁切换完全不同的任务，比如先分析代码、再处理图片、再写文案等，新的问题会不断引入不同的上下文，导致原本可复用的缓存被重写或命中率下降。建议把不同类型的任务拆到不同会话里处理，例如代码分析、图片处理、文案创作分别开新会话，这样更有利于缓存复用。\n","id":6,"question":"为什么感觉消耗越来越快？如何减少额度消耗"},{"answer":"- 401：密钥缺失、无效，或认证来源冲突。\n- 403：余额、令牌分组、模型权限或策略拒绝。\n- 404：Base URL、/v1/responses 路径或模型 ID 不匹配。\n- 429：降低并发，按服务端提示退避重试。\n- 5xx：保留请求 ID，短暂退避后重试一次。\n- Connection failed：先用 curl 验证同一 URL；再检查 DNS、代理、证书和 IDE 启动环境。\n\n不要因为其他终端工具能联网就直接断定是 MTU。只有在抓包或可重复的包大小测试证明后才调整网络参数。","id":7,"question":"401、403、404 和网络错误怎么排查？"},{"answer":"如果你遇到以下情况，可以联系我们排查：\n- 明显没有发起任务，但账单持续产生记录。\n- 短时间内多次重复 Cache 写入，并且没有 Cache 读取。\n- 同一请求出现异常大量重复扣费。\n- 账单模型与你的使用场景完全不符。\n- 实扣金额明显异常。\n- Agent 调用失败，但仍产生了无法理解的扣费记录。\n请进群添加客服，尽量提供请求时间、模型名称、来源和截图，方便我们快速定位。","id":8,"question":"什么时候需要联系客服？"}],"faq_enabled":true,"footer_html":"\u003cp style=\"font-size: 14px; margin-top: 10px;\"\u003e© 2026 WolfAI 版权所有 联系我们：support@wolfai.top\u003c/p\u003e","github_client_id":"73f9239a4b2387a4276f","github_oauth":true,"linuxdo_client_id":"","linuxdo_minimum_trust_level":0,"linuxdo_oauth":false,"log_export_enabled":true,"log_max_range_days":365,"logo":"https://mj.hdgsb.com/wolflogo.png","mj_notify_enabled":true,"oidc_authorization_endpoint":"","oidc_client_id":"","oidc_display_name":"OIDC","oidc_enabled":false,"passkey_allow_insecure":false,"passkey_attachment":"","passkey_display_name":"WolfAI","passkey_login":true,"passkey_rp_id":"wolfai.top","passkey_rp_ids":["wolfai.top","api.hdgsb.com","www.wolfai.top"],"passkey_user_verification":"preferred","password_login_enabled":true,"password_login_encryption_enabled":false,"password_register_enabled":true,"price":2,"privacy_policy_enabled":true,"quota_display_type":"USD","quota_per_unit":500000,"quota_remind_enabled":true,"quota_remind_threshold":1000,"register_enabled":true,"self_use_mode_enabled":false,"server_address":"https://wolfai.top","setup":true,"start_time":1794304978,"stripe_unit_price":2,"system_name":"WolfAI - 全球AI聚合平台","telegram_bot_name":"","telegram_oauth":false,"telegram_oauth_configured":false,"theme":"default","turnstile_check":false,"turnstile_site_key":"0x4AAAAAAC95r64n6-0o-tZj","uptime_kuma_enabled":true,"usd_exchange_rate":1,"user_agreement_enabled":true,"version":"v1.0.0-rc.49","wechat_login":true,"wechat_qrcode":"https://mj.hdgsb.com/zzh.jpg"},"message":"","success":true}