nvidia/AceReason-Nemotron-14B

The nvidia/AceReason-Nemotron-14B is a 14 billion parameter language model developed by NVIDIA, specifically designed for advanced math and code reasoning. Trained entirely through reinforcement learning, starting from DeepSeek-R1-Distilled-Qwen-14B, it achieves strong performance on benchmarks like AIME and LiveCodeBench. This model excels at solving complex mathematical problems and generating accurate code, leveraging its 32768 token context length.