ChinaGenAIFree API key
Compare API accessAll guidesWhat’s new
Search
OpenAI-compatible Chat Completions

Free OpenAI-compatible API

One limited free Chat Completions endpoint for compatible apps, scripts and small tools. Model routing is automatic.

Check key accessRead the quickstart
Checking site API configuration… Service status · Compare separate API routes
Your connectionOpenAI-compatible Chat Completions
Checking
API base URL
https://chinagenai.com/v1
API key
YOUR_SITE_KEYCheck access
Model
freeAutomatic routing
curl 'https://chinagenai.com/v1/chat/completions' \
  -H 'Authorization: Bearer YOUR_SITE_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "free",
    "messages": [{"role": "user", "content": "Hello!"}],
    "max_tokens": 128
  }'
Example only. Replace YOUR_SITE_KEY before use.
Already have a response? Inspect JSON or SSE locally →
Local tools without sign-in Clear usage limits
For small text tasks256 output tokens1 active request / key20 attempts / dayCheck your task fit →

A familiar API.
A shorter setup.

Check access, then connect with three settings.

  1. 1

    Check free key access

    When available, verify your browser and copy a site key. No email or password.

  2. 2

    Connect your tool

    Paste the API base, key and free model label into your app.

  3. 3

    Send a short request

    One active request per key. View your allowance in the console.

Start with a model you know.

Explore provider guides, free-access conditions and practical setup.

Search exact model IDsCompare API accessModel guidesFree Chinese model APIsGLM API and Coding Plan
Provider / routeAccess & limitsStart here
D

DeepSeek

DeepSeek API
Metered API

Current V4.1 Flash model name: deepseek-flash. Direct API use is metered; free consumer chat access is separate.

API access & setupDeepSeek to valid JSON
Q

Qwen

Alibaba Cloud Model Studio
Check trial eligibility

Trial eligibility, expiry and paid overage depend on the provider and account. Check your actual quota first.

API access & setupFree Qwen API: daily allocation and trial checks
O

OpenRouter

Multi-provider routing
Specific free variants

Only models with a real free catalog entry qualify. A :free suffix does not make any model free.

API access & setupOpenRouter free models: IDs and image inputs
K

Kimi

Moonshot AI · China and international Kimi API
¥10 China / $1 international minimum; metered

China site: ¥10 minimum top-up. International site: $1 minimum. Accounts, keys and balances are separate; K2 access and shutdown dates are approaching.

API access & setup
M

Xiaomi MiMo

Xiaomi MiMo API
Metered API; Token Plan is separate

Current standard text routes include mimo-v2.6-pro and mimo-v2.6-flash. Xiaomi lists mimo-v2.6-pro-ultraspeed as a customized service. The pay-as-you-go API and Token Plan use different endpoints and credentials.

API access & setuptool_calls empty content 400: check request history
D

Doubao (Volcengine Ark)

Volcengine Ark · China (Beijing)
Paid API; limited trial may apply

Eligible accounts may get a limited safe-trial quota for models that are not enabled. Verify the model, region and remaining quota in Ark.

API access & setupDoubao API setup: base URL, model ID and trial quota
6 of 6 routes · Review dates appear on each provider guide. No performance ranking implied.

These are independent provider guides. Our free API routes automatically and does not guarantee any named model. See how we check access and describe these guides.

Make the connection useful.

A little guidance between an API key and a working tool.

Diagnose an API response
Browse guides
5 min read

Free API for Cherry Studio

Set the API Host, site key and free model label. Resolve expired keys, rate limits and short answers.

Read guide
5 min read

JSON output cut off? Diagnose finish_reason: length

Inspect an incomplete or cut-off JSON response, finish_reason, syntax and missing fields without making another API call.

Read guide
6 min read

Connect a chat API to n8n

Choose an n8n node and API, route a local model from Docker, diagnose empty model lists and AI Agent v2 model ID errors, and prepare a JSON request.

Read guide

Build a request. Check the JSON.

Local tools for preparing a request and checking structured output. Your text stays in the browser.

Open the workbench

A few things to know.

Simple limits, clear expectations.

Full connection guide
Which software can I connect?

Apps and scripts that accept a custom OpenAI-compatible Chat Completions URL and API key. Check the connection guide for supported features. Some apps require endpoints that this service does not provide.

Do I need to choose a model?

Use free as the model label. The service selects its upstream route. Provider names in our guides help you explore options; they do not select or guarantee a specific model here.

What does the free allowance include?

When enabled, 20 attempts per network or verified account per UTC day, 48-hour keys and one active request per key. The site shares two active slots and a daily request cap. Shared Wi-Fi shares allowance. Requests accepted for processing count even if the upstream call fails.

Is a key available right now?

The notice reports whether the site API configuration is enabled; it does not check remaining quota or upstream availability. Open API access for key access and controls. If you need a different endpoint, compare separate API routes.