DeepSeek-V3
by DeepSeek model
Superseded. DeepSeek's current general-purpose model is deepseek-v4-flash, with a 1M-token context window and up to 384K tokens of output. V3 is still very widely referenced and remains one of the most commonly self-hosted open-weight models, which is why this page is kept rather than removed.
Visit siteGood at
- Chat 3 / 5
- Code Generation 3 / 5
- Text Generation 3 / 5