Phi-4-Reasoning

Phi-4-Reasoning is a 14-billion parameter open-weight reasoning model from Microsoft, fine-tuned from Phi-4 using supervised fine-tuning on chain-of-thought datasets. Despite its small size, it rivals much larger models on complex reasoning tasks, performing on par with OpenAI's o1-mini and o3-mini. It uses custom thinking tags to structure logical blocks and can be extended up to 64k tokens. Also available as Phi-4-Reasoning-Plus with enhanced reasoning and Phi-4-Mini-Reasoning for edge deployment.