nvidia/AceMath-RL-Nemotron-7B

The nvidia/AceMath-RL-Nemotron-7B is a 7.6 billion parameter math reasoning model developed by NVIDIA, trained entirely through reinforcement learning (RL) from Deepseek-R1-Distilled-Qwen-7B. It achieves 69.0% Pass@1 accuracy on AIME 2024 and 53.6% Pass@1 accuracy on AIME 2025, demonstrating strong performance in advanced mathematical problem-solving. This model also shows improved coding accuracy on LiveCodeBench, reaching 44.4% Pass@1, indicating generalization capabilities from its RL training. It is optimized for complex mathematical reasoning and can also be applied to coding tasks.