The Sequence Research #533: NVIDIA's Nemotron Models and the OpenMathReasoning Dataset Kill it in the AI Math Olympiad
The new models advanced the frontiers of mathematical reasoning.
If you follow this newsletter, you know I am obsessed with AI in math. Today, I would like to discuss one of the newest advacements in this area.
In April 2025, NVIDIA reached a significant milestone in AI by securing first place in the AI Mathematical Olympiad – Progress Prize 2 (AIMO-2). This achievement was driven by two advanced large language models: OpenMath-Nemotron-32B and OpenMath-Nemotron-14B-Kaggle. These models, trained on the OpenMathReasoning dataset, establish new state-of-the-art capabilities in mathematical reasoning.