nvidia/OpenMath2-Llama3.1-8B

OpenMath2-Llama3.1-8B is an 8 billion parameter language model developed by NVIDIA, fine-tuned from Llama3.1-8B-Base with the OpenMathInstruct-2 dataset. This model is specifically optimized for mathematical reasoning and problem-solving, demonstrating significant performance improvements over Llama3.1-8B-Instruct on various math benchmarks, including a 15.9% increase on the MATH dataset. It is designed for advanced mathematical tasks, leveraging a 32768 token context length.