DeepSeek-Math 7B
An open-source 7-billion-parameter mathematical large language model released by DeepSeek, focused on mathematical reasoning and problem solving. Its reinforcement-learning variant, combined with tool invocation, achieves approximately 60% accuracy on the MATH benchmark, offering a notable capability-to-cost ratio among models of comparable scale.
Timeline (last 90 days)
DeepSeek-Math 7B is an open-source mathematics large model with 7B parameters, whose RL version with tool invocation can achieve approximately 60% accuracy on the MATH benchmark, approaching GPT-4 level.
DeepSeek-Math provides three versions on Hugging Face: a Base pretrained version, an Instruct dialogue version, and an RL reinforcement learning version.
All Facts (2)
DeepSeek-Math 7B is an open-source mathematics large model with 7B parameters, whose RL version with tool invocation can achieve approximately 60% accuracy on the MATH benchmark, approaching GPT-4 level.
50%UnverifiedDeepSeek-Math provides three versions on Hugging Face: a Base pretrained version, an Instruct dialogue version, and an RL reinforcement learning version.
50%