Mathstral 7B
Mistral's 7B math-specialized model with 32K context for STEM reasoning.
About
Mathstral 7B is Mistral AI's math-specialized 7-billion-parameter model, released on July 16, 2024 in collaboration with Project Numina. Built on the Mistral 7B architecture and further trained on mathematical and scientific data, it targets multi-step logical reasoning in STEM fields with a 32K-token context window for long reasoning chains.
It scored 56.6 percent on the MATH benchmark at release, roughly double the base Mistral 7B, with further gains from majority voting and reward-model reranking. Weights are released under Apache 2.0 on Hugging Face, making it a standard open baseline for math reasoning research.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 130; Trending score 80/100 from researcher assessment.