How to get a free DeepSeek API key in 2026
DeepSeek's own API is paid. Here is how to call DeepSeek models for free through an OpenAI compatible key, what the limits are, and how to use it in your code, Janitor AI or SillyTavern.
DeepSeek's own API is paid. To use DeepSeek for free, create a UnoRouter account, copy the API key and call a model ending in :free, such as deepseek-v4.1-flash-1017:free, at https://api.unorouter.com/v1. The free tier allows about one request per minute per model.
DeepSeek's official platform bills every token, so there is no free API key from DeepSeek itself. You can still call DeepSeek models for free: UnoRouter serves them on a free tier through one OpenAI compatible key. Here is how to get that key in two minutes and what you can do with it.
Is the DeepSeek API free?
Not from DeepSeek directly: its platform is pay per token. Free access comes from gateways that run DeepSeek models on free capacity. On UnoRouter, every model id ending in :free costs nothing, and at the time of writing that includes deepseek-v4.1-flash-1017:free, deepseek-v4-pro:free and deepseek-chat:free. Free models come and go with capacity, so the models page shows what is free right now.
Get the key in two minutes
Create an account, the fastest way is signing in with Google, Discord or GitHub. Open the Tokens page and copy your API key. That is the whole setup: no card, no top up. The key works with any OpenAI compatible library or client.
Call DeepSeek with it
Set the base URL to https://api.unorouter.com/v1, use your key as the Bearer token and pick a free DeepSeek model, for example model: deepseek-v4.1-flash-1017:free on POST /v1/chat/completions. Streaming and tool calls work the same way as with OpenAI's API, so existing code needs only the base URL and model name changed.
The limits
Free models allow about one request per minute per model and account. Go faster and you get HTTP 429 with a Retry-After header: wait that long, or switch to another free model, which has its own allowance. The docs list every limit and error code. For steady traffic, paid DeepSeek models on the same key start at about $0.16 per million input tokens.
Use it in your apps
The same key works for coding agents like Cline or OpenCode, for roleplay in Janitor AI and SillyTavern, and in the built in chat. You do not need a separate key per app or per model.
Get your free key and see which DeepSeek models are free right now.
Frequently asked questions
Is the DeepSeek API free?
Not on DeepSeek's own platform, which bills per token. UnoRouter serves DeepSeek models for free through any model id ending in :free.
How do I get a free DeepSeek API key?
Sign up at UnoRouter with Google, Discord or GitHub, copy your key from the Tokens page, and use it with the base URL https://api.unorouter.com/v1 and a free DeepSeek model. No credit card is needed.
What are the limits of the free DeepSeek API?
About one request per minute per model and account. Faster requests get HTTP 429 with a Retry-After header; wait or switch to another free model.
Trending models this week
Chat, image, embedding and speech models you can call for free through one OpenAI compatible API, what the limits are, and how to start in two minutes.
What a proxy does in Janitor AI, what makes one good for roleplay, and how the free and paid options compare, including where UnoRouter fits and where it does not.
Wearing the UnoRouter Discord tag shortens your free-model rate limit window by 25 percent, so a model you could call once a minute you can call every 45 seconds.
Comments
Loading comments