OpenMath-Nemotron-32B Overview

OpenMath-Nemotron-32B is a 32.8 billion parameter model from NVIDIA, built upon the Qwen2.5 architecture. It is specifically fine-tuned using the OpenMathReasoning dataset to excel in complex mathematical reasoning. This model demonstrates state-of-the-art performance across various mathematical benchmarks, as detailed in its accompanying paper.

Key Capabilities

• Advanced Mathematical Reasoning: Achieves high accuracy on benchmarks like AIME24, AIME25, HMMT-24-25, and HLE-Math, particularly when utilizing advanced inference modes.
• Flexible Inference Modes: Supports Chain-of-Thought (CoT), Tool-Integrated Reasoning (TIR), and Generative Solution Selection (GenSelect) for diverse problem-solving approaches.
• Long Context Window: Features a substantial context length of 131,072 tokens, enabling the processing of extensive mathematical problems and contexts.
• Commercial Use Ready: Licensed under CC-BY-4.0 and Apache License Version 2.0, making it suitable for commercial applications.

Good For

• Mathematical Problem Solving: Ideal for applications requiring precise and advanced mathematical reasoning.
• Research in AI for Math: Provides a robust foundation for further research and development in automated mathematical problem-solving.
• Competitive Math Challenges: A version of the 14B model secured first place in the AIMO-2 Kaggle competition, highlighting its capability in competitive environments.

It is important to note that this model is not instruction-tuned for general data and is optimized specifically for the mathematical domain.