DeepSeek models
Frontier reasoning under MIT - V4 online, Flash and R1 to run yourself.
Run on your device
DeepSeek V4 Flash
DeepSeek · Jul 2026DeepSeek's V4 Flash - a 284B mixture-of-experts with 13B active per token and a 1M-token context. Frontier-class reasoning and agentic work on a workstation with a big graphics card and 128 GB or more of main memory.
from 96.9 GBruns from 128 GB RAM1M contextMIT
DeepSeek R1 0528
DeepSeek · May 2025Powerful chain-of-thought reasoning. Excels at complex problem solving, math, and logic.
from 5.0 GBruns from 16 GB RAM128K contextMIT