What it is

DeepSeek is the AI assistant built by DeepSeek-AI on top of the fully open-source DeepSeek-V3 and R1 models. R1’s reinforcement-learning reasoning makes it stand out on math, code and logic, while its API costs a fraction of comparable international models.

Who it is for

Students, writers and developers alike. Use the free web or app versions for daily Q&A and writing; integrate via API for products at negligible cost; or self-host the open weights with vLLM or Ollama.

How to use

Open the website or app and start chatting — enable “DeepThink” to invoke R1 reasoning. Developers can grab an API key from the open platform; the interface is OpenAI-compatible, so migration takes one line.

Pricing

Web and app are free. API usage is billed per token with discounted cached input. Check the official platform for current rates.

FAQ

R1 vs normal mode? R1 shows its chain of thought before answering — slower but markedly more accurate on hard problems.

Can I use the open weights commercially? Yes, the weights are released under the MIT license.