Key Takeaways
- Mistral 7B scores 62.5% on MMLU benchmark
- Mixtral 8x7B achieves 70.6% on MMLU (5-shot)
- Mistral Large v0.1 scores 81.2% on MMLU
- Mistral AI was founded in April 2023 by Arthur Mensch, Guillaume Lample, and Timothée Lacroix
- Mistral AI raised €105 million in seed funding in June 2023 at a €260 million post-money valuation
- Mistral AI secured $415 million in Series A funding in June 2024 led by General Catalyst
- Mistral 7B model has 7.3 billion parameters
- Mixtral 8x7B is a sparse Mixture of Experts model with 46.7B total parameters but 12.9B active
- Mistral Large v0.1 has undisclosed parameters but outperforms GPT-4 on most benchmarks
- Mistral AI partnered with Nvidia for GPU optimization
- Microsoft invested in Mistral and hosts models on Azure AI
- AWS integrated Mistral models into Bedrock platform
- Mistral AI's Le Chat app has over 5 million monthly active users as of 2024
- Mistral models have been downloaded over 100 million times on Hugging Face
- Mistral AI's API serves over 1 million requests per day
Mistral’s models deliver standout benchmarks and fast real world adoption, from 81.2% MMLU to million user Le Chat.
Related reading
01 · Category
Benchmark Performance23 stats
Benchmark Performance Interpretation
02 · Category
Company Founding24 stats
Company Founding Interpretation
03 · Category
Model Architecture25 stats
Model Architecture Interpretation
04 · Category
Partnerships17 stats
Partnerships Interpretation
05 · Category
User Adoption21 stats
User Adoption Interpretation
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
Christopher Morgan. (2026, February 24). Mistral AI Statistics. Gitnux. https://gitnux.org/mistral-ai-statistics
Christopher Morgan. "Mistral AI Statistics." Gitnux, 24 Feb 2026, https://gitnux.org/mistral-ai-statistics.
Christopher Morgan. 2026. "Mistral AI Statistics." Gitnux. https://gitnux.org/mistral-ai-statistics.
Sources & references
41 datasets cited across this report · attribution is report-level

