189 7 months ago

(Unsloth Dynamic Quants) DeepSeek's first-generation of reasoning models with comparable performance to OpenAI-o1, 671B MoE full model, not dense distilled models.