DeepSeek
DeepSeek is a Chinese AI lab best known for releasing frontier-class models as open weights. You can use them through the DeepSeek chat app (web, iOS, Android and desktop), call them through the API, or download them and run them yourself.
What it does now
- The chat app runs on DeepSeek-V4.1-Flash, which brings better text and agent performance and native image understanding.
- Open weights under the MIT license. V4.1-Flash has 552B backbone parameters and supports contexts up to one million tokens, according to its Hugging Face model card.
- DeepSeek-V4-Pro is the larger model, available through the API.
API pricing (checked 15 Oct 2026 on api-docs.deepseek.com)
Prices are per million tokens. Off-peak rates are half the peak rate.
| Model | Input (cache hit) | Input (cache miss) | Output |
|---|---|---|---|
deepseek-flash (off-peak / peak) | $0.003 / $0.006 | $0.15 / $0.30 | $0.60 / $1.20 |
deepseek-v4-pro (off-peak / peak) | $0.022 / $0.044 | $0.66 / $1.32 | $1.98 / $3.96 |
Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday.
We couldn’t find a paid subscription for the chat app.
Things to know
- Check DeepSeek’s privacy policy for where your data is stored and how it’s used before sharing anything sensitive. Running the open weights yourself avoids the question entirely.
Who it’s for
- Developers who want frontier-class models at a fraction of the usual API price.
- Teams that need to self-host a strong model under a permissive license.