Is DeepSeek free? Chat is, the API costs from $0.15 per 1M tokens

Yes, DeepSeek is free to use as a chat assistant, but its API is not. DeepSeek's own site describes its official assistant as free to try, and the API pricing page charges per token. This is what each costs, as of October 27, 2026.

The two get mixed up because they are separate products. The chat app and website are for people. The API is for developers who put DeepSeek models inside their own software, and it is billed by usage.

Is the DeepSeek chat app free?

DeepSeek describes its official assistant as a free way to talk to its latest models. That wording comes from the app description on deepseek.com, which is in Chinese: it says users can try the assistant for free. We could not open chat.deepseek.com itself, which blocked automated requests with a 403 error on October 27. So we can't tell you from the chat page what limits apply, whether any features are held back, or whether that changes.

DeepSeek's pages that we read did not list a paid chat plan or a message cap. If you want to confirm the current terms, sign in at chat.deepseek.com and check there.

Is the DeepSeek API free?

No. The API charges for tokens, the small chunks of text a model reads and writes. Prices are per 1 million tokens, in US dollars, and DeepSeek lists two models. Fees are taken from a topped-up balance or a granted balance, and the granted balance is used first when you have both. DeepSeek does not say on that page how large any granted balance is.

DeepSeek API pricing

Prices below are from DeepSeek's pricing page, as of October 27, 2026. DeepSeek says it may change prices.

A cache hit means the start of your prompt matches one DeepSeek has recently processed, so it costs less to read again. A cache miss is a prompt it has to read fresh.

Both models have a 1M-token context window and can write up to 384K tokens in one reply. deepseek-flash is served by DeepSeek-V4.1-Flash-1017 and accepts images. deepseek-v4-pro is DeepSeek-V4-Pro-0813 and does not. The older names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work but are billed at the Flash price.

Peak and off-peak hours

Off-peak rates are half of the peak rates. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, except Chinese public holidays. Every other time, including weekends, is off-peak. If your job can wait, running it off-peak halves the bill.

Search and fetch cost extra

DeepSeek's API also offers web search and page fetch. Each costs $0.75 per 1,000 successful requests, as of October 27, 2026. Calls made through DeepSeek Harness, Claude Code or Codex are free for now. DeepSeek says per-request billing starts at 00:00 Beijing time on November 1, 2026.

What to do next

If you only want to chat, use the app or website. If you are building something, start with deepseek-flash, which is the cheaper model, and check the pricing page before you commit to a budget.