AI API GATEWAY

Top models, OpenAI-compatible. 主流模型，OpenAI 兼容中转。

Hachimi Cloud relays Claude, GPT & Gemini in OpenAI-compatible format — each provider on its own channel. Carpool the cost, keep native tools working out of the box. 哈基米云以 OpenAI 兼容格式中转 Claude、GPT、Gemini —— 每家模型各走独立渠道。拼车共享、分摊成本，原生工具开箱即用、无缝直连。

OpenAI-compatible GPT · GLM Carpool sharing拼车共享

Get started立即开始 Read the docs查看文档

01Available Models可用模型

GPT — OpenAI models.GPT — OpenAI 模型。

Model	Context上下文	Output输出
`gpt-5.4-mini` GPT-5.4-Mini	400K	128K
`gpt-5.6-luna` GPT-5.6-Luna	400K	128K
`gpt-5.6-terra` GPT-5.6-Terra	400K	128K
`gpt-5.6-sol` GPT-5.6-Sol	400K	128K

Context limits vary by model. All GPT models pass store: false to prevent API-side conversation storage.上下文上限因模型而异。所有 GPT 模型均设置 store: false 以阻止 API 端存储对话。

GLM — Zhipu AI reasoning models.GLM — 智谱 AI 推理模型。

Model	Context上下文	Output输出
`glm-5.2` GLM-5.2	1M	128K

Claude — Anthropic models.Claude — Anthropic 模型。

Model	Context上下文	Output输出
`claude-haiku-4-5` Anthropic	200K	64K
`claude-sonnet-5` Anthropic	1M	128K
`claude-opus-4-8` Anthropic	1M	128K
`claude-fable-5` Anthropic	1M	128K

Full model list lives in the docs.完整模型列表见文档。

02Pricing定价

Pay per token in CNY (¥), metered in real time. Prices shown per 1M tokens.按 token 实时计费，单位 人民币元（¥），下表价格均为每 100 万 tokens（1M）。

Model	Input输入	Output输出	Cache write缓存写入	Cache read缓存读取
`gpt-5.4-mini`	¥0.15	¥0.9	¥0	¥0.015
`gpt-5.6-luna`	¥0.2	¥1.2	¥0.25	¥0.026
`gpt-5.6-terra`	¥0.5	¥3	¥0.625	¥0.065
`gpt-5.6-sol`	¥1	¥5	¥1.25	¥0.13

Model	Input输入	Output输出	Cache write缓存写入	Cache read缓存读取
`glm-5.2`	¥8	¥28	—	¥2

Model	Input输入	Output输出	Cache write缓存写入	Cache read缓存读取
`claude-haiku-4-5`	¥1	¥5	¥1.25	¥0.1
`claude-sonnet-5`	¥2	¥10	¥2.5	¥0.2
`claude-opus-4-8`	¥5	¥25	¥6.25	¥0.5
`claude-fable-5`	¥10	¥50	¥12.5	¥1

The GPT-5.6 series charges for cache writes (GPT-5.4-Mini does not bill cache writes separately). Final rates may vary by user group — see your console for live quotas.gpt-5.6 系列已正式收取缓存写入费用（gpt-5.4-mini 暂不单独计费）。实际倍率可能因用户组而异，实时额度以控制台为准。

03Why Hachimi Cloud为什么选哈基米云

OpenAI-compatible relayOpenAI 兼容中转

Reach Claude, GPT & Gemini in OpenAI-compatible format — each provider runs as its own channel. Point Claude Code, Codex or OpenCode at it and native tools just work.以 OpenAI 兼容格式访问 Claude、GPT、Gemini —— 每家模型对应独立渠道。Claude Code、Codex、OpenCode 直连即用，原生工具无缝可用。

Carpool & pooling拼车 · 账号池

Multi-account load balancing with sticky sessions. Share subscriptions, split the bill, and stay efficient.多账号负载均衡 + 粘性会话。订阅拼车共享，分摊成本，更高效。

Realtime billing实时计费配额

Usage metered in real time with transparent quotas — see exactly where every token goes.用量实时计费，额度透明可查 —— 每个 token 的去向一目了然。

04Works with your tools兼容你的工具

Drop-in for the agents and providers you already use.无缝接入你已经在用的 Agent 与模型厂商。

CCClaude Code CXCodex OCOpenCode CSCC Switch

CClaude GGPT GGemini AAntigravitysoon即将

05Quick start快速开始

Wire up Claude Code in one config — swap the key and model, you're live.一段配置接入 Claude Code —— 换上你的密钥和模型即可使用。

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-your-api-key-here",
    "ANTHROPIC_BASE_URL": "https://api.hachimi.cloud",
    "ANTHROPIC_MODEL": "gpt-5.5"
  }
}

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-your-api-key-here",
    "ANTHROPIC_BASE_URL": "https://api.hachimi.cloud",
    "ANTHROPIC_MODEL": "glm-5.2"
  }
}

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-your-api-key-here",
    "ANTHROPIC_BASE_URL": "https://api.hachimi.cloud",
    "ANTHROPIC_MODEL": "claude-sonnet-5"
  }
}

OpenAI-compatible — point any OpenAI client at https://api.hachimi.cloud/v1 with your key. OpenCode & Codex setup lives in the docs. OpenAI 兼容 — 任意 OpenAI 客户端把 base_url 指向 https://api.hachimi.cloud/v1 并填入密钥即可。OpenCode、Codex 配置见文档。

Create your key创建你的密钥