nvidia/OpenMath-Nemotron-14B-Kaggle

The OpenMath-Nemotron-14B-Kaggle model by NVIDIA is a 14.8 billion parameter, Qwen2.5-based transformer decoder-only language model, fine-tuned on a subset of the OpenMathReasoning dataset. With a 131,072 token context length, it is specifically optimized for advanced mathematical reasoning and problem-solving, achieving strong results on benchmarks like AIME and HMMT. This model was instrumental in NVIDIA's first-place submission to the AIMO-2 Kaggle competition, demonstrating its capability in complex mathematical tasks.