Back to Deepseek V3
DeepSeek's 671B MoE chat model — frontier-level instruction following at open-weight cost, rivalling GPT-4o on most benchmarks.
164K tokensMoEProprietary
Benchmarks
DeepSeek-V3
76.7%DeepSeek-V3-0324
76.4%DeepSeek Reasoner (V3.2)
72.8%DeepSeek Chat (V3.2)
39.5%DeepSeekV3.2 agent submission