Best LLMs for Science — Mistral
The best Mistral model for science is Mistral Medium 3.5, scoring 74.8% on GPQA Diamond — graduate-level physics, chemistry and biology. Mistral Medium 3.1 (58.8%) and Mistral Medium 3 (57.8%) round out the top three.
- 1
mistral-medium-3-5ReasoningToolsJSON+129.9 intel · $1.50/M · 152 t/s74.8%GPQA - 2mistral-medium-3.1ToolsJSONVision14.7 intel · $0.400/M · 45 t/s58.8%GPQA
- 3mistral-medium-3ToolsJSONVision12.5 intel · $0.400/M · 41 t/s57.8%GPQA
- 4mistral-large-2407ToolsJSON7.3 intel · $2.00/M · 131K ctx47.2%GPQA
- 5mistral-sabaToolsJSON6.4 intel · $0.200/M · 33K ctx42.4%GPQA
- 6mistral-largeToolsJSON4.4 intel · $2.00/M · 128K ctx35.1%GPQA
- 7mixtral-8x22b-instructToolsJSON4.4 intel · $2.00/M · 66K ctx33.2%GPQA
How this is ranked
AI models ranked by GPQA Diamond — graduate-level physics, chemistry and biology questions written to be un-Googleable, so the score reflects reasoning from knowledge rather than retrieval. The best large language models for scientific research and hard-science work.
Frequently asked
What is the best Mistral model for science?
The best Mistral model for science is Mistral Medium 3.5, scoring 74.8% on GPQA Diamond — graduate-level physics, chemistry and biology. Mistral Medium 3.1 (58.8%) and Mistral Medium 3 (57.8%) round out the top three.
What's a good alternative to Mistral Medium 3.5?
Mistral Medium 3.1 (58.8%) is the closest alternative on this metric, followed by Mistral Medium 3 (57.8%). See the full ranking above for the tradeoffs.
How many Mistral models are there?
modelgrep tracks 18 Mistral models with live benchmarks, speed, latency and per-provider pricing, led on intelligence by Mistral Medium 3.5. 7 of them qualify for this ranking.