Flagship
A 552B mixture of experts built as a causal encoder and decoder, with 8B parameters active per token in prefill and 16B in decode.
Read: DeepSeek-V4.1-FlashDeepSeek
Hangzhou, ChinaFounded 2023deepseek.com
DeepSeek publishes the weights of its current models under the MIT licence. The largest, DeepSeek-V4-Pro, has 1.6T parameters.
Flagship
Lineup
DeepSeek-V4.1-Flash
Flagship
- Total
- 552B
- Active
- 16B
- Context
- 1M
- Total
- 1.6T
- Active
- 49B
- Context
- 1M