FractalAIResearch/Fathom-R1-14B

FractalAIResearch's Fathom-R1-14B is a 14-billion-parameter reasoning language model derived from Deepseek-R1-Distilled-Qwen-14B, optimized for mathematical reasoning within a 16K context window. It achieves performance comparable to closed-source o4-mini models on olympiad-level math exams like AIME-25 and HMMT-25, with a total post-training cost of $499. This model excels at complex mathematical problem-solving, demonstrating significant improvements over its base model and other open-source alternatives.