Hugging Face Overview
OpenReasoning-Nemotron-14B Overview
OpenReasoning-Nemotron-14B, developed by NVIDIA, is a 14.8 billion parameter large language model derived from Qwen2.5-14B. It is specifically post-trained to excel in reasoning across math, code, and science solution generation, supporting an extensive context length of 131072 tokens. The model demonstrates strong performance on challenging reasoning benchmarks, with the 14B variant achieving scores like 60.9 on Artificial Analysis Index, 71.6 on GPQA, and 77.5 on MMLU-PRO.
Key Capabilities
- Advanced Reasoning: Optimized for complex problem-solving in mathematics, programming, and scientific domains.
- GenSelect Integration: Can be used with a "heavy" inference mode called GenSelect, which combines multiple parallel generations to significantly improve performance on math and coding benchmarks, often surpassing pass@1 scores.
- High Output Token Support: Capable of generating solutions up to 64,000 output tokens, suitable for detailed problem-solving.
- Commercial Use: Available for both commercial and non-commercial research applications under the Creative Commons Attribution 4.0 International License (CC-BY-4.0).
Good For
- Developers and researchers focused on competitive math, code, and science problems.
- Applications requiring robust reasoning capabilities and detailed solution generation.
- Leveraging advanced inference techniques like GenSelect for enhanced accuracy in problem-solving.