Want to use a free API in Cherry Studio without writing code? Add ChinaGenAI as a custom provider for short text chats. You need a ChinaGenAI site key; an OpenAI, DeepSeek or Qwen account key is a different credential.
Verify your browser to get a 48-hour key. No email, payment or ad click is required. Get a free API key.
Cherry Studio settings for ChinaGenAI
- Provider name
- ChinaGenAI — custom provider
- OpenAI API Host / Base URL
https://chinagenai.com- API key
- Your site key from API access
- Model ID
free- Expected request URL
https://chinagenai.com/v1/chat/completions
The free label selects the site's automatic route. It does not select a specific DeepSeek, Qwen, Kimi or overseas model.
Add the provider in Cherry Studio desktop
- Open Settings → Model Provider → + Add Provider. Name it ChinaGenAI and use the OpenAI endpoint field.
- Enter the host above and your site key. Check the request-path preview: it should match the expected URL, with one
/v1. - Save the provider. Add model ID
freemanually, or sync models and add it from the returned list. Enable the provider. - In ordinary Chat, select ChinaGenAI / free. Start a fresh, short conversation when you are ready to use the allowance.
For a version-style host field, https://chinagenai.com/v1/ is the documented alternative: Cherry Studio appends chat/completions. Do not paste the full endpoint into a field that appends that path again. Use the displayed request URL to resolve differences between app versions.
Official custom-provider instructions
Official API Host and model settings
Will this free route fit your task?
Use it for a short rewrite, a brief answer or a small structured result. It allows 20 attempts per network per UTC day, 1 active request per key and 2 across the site. The whole site also shares 200 attempts per UTC day. People on the same Wi-Fi share the network allowance.
Each request supports up to 6 messages and 6,000 serialized input characters, including instructions and tool definitions. Output is capped at 256 tokens. A long chat history can exceed the input limit even when your newest message is short. A fresh topic with only the necessary text is a better fit.
For example, you could ask for one short rewrite of a paragraph. This is an instruction example, not a recorded model result:
Rewrite this sentence in plain English, in under 20 words: Please advise whether the requested item is currently available.Choose ordinary text Chat. Cherry Agent requires an Anthropic-compatible connection; this site exposes Chat Completions. Knowledge-base embeddings, image generation, audio and OpenAI Responses are outside this endpoint. Streaming is buffered: the answer can arrive after a wait rather than appearing token by token.
Official provider protocol types
Free API not working? Separate these cases
Check the HTTP status and error body if your client exposes them. The cases below describe ChinaGenAI's gateway responses; Cherry Studio's wording can vary by version.
| Signal | What to do |
|---|---|
Answer cuts off; finish_reason: length | The output reached its limit. A valid key can still return an incomplete answer. Ask for a smaller result; requesting more than 256 output tokens does not raise this site's cap. |
401: key rejected | Check the site key, including its 48-hour expiry. Replace it in the console if needed, then update Cherry Studio's saved credential. Changing the model name does not repair a key. |
429: daily or concurrent limit | Let active requests finish. Daily allowance resets at 00:00 UTC; replacing a key does not reset it. This route rejects excess requests instead of placing them in a queue. |
400 or an unexpected path | Compare the request URL with the settings above. Check message count, input size and whether the app is requesting a supported chat endpoint. |
502 or 503 | The upstream failed or the service is unavailable. Read API status before retrying. An upstream attempt can count toward allowance even when it fails. |
If you have a non-streaming JSON response, paste it into the local response inspector to check the finish reason and JSON syntax. It makes no model call. For another client, use the general OpenAI-compatible connection guide.
Updated 19 October 2026 from the linked Cherry Studio documentation and this site's implemented limits. The configuration is documentation-based; a live Cherry Studio session has not been verified here. If these limits are too small for your task, use a provider whose published service fits it.
Free API access and limits
Check the site free API access and limits before trying to obtain a site key. Read the current status and request limits before using a site key. It is not a source of official provider keys.
ChinaGenAI is not affiliated with the providers named here. This release contains setup guidance and explicitly labeled fictional examples, not claimed live model benchmarks.