# Mistral AI
French AI company headquartered in Paris. Founded in April 2023 by [[Arthur Mensch]], [[Guillaume Lample]], and [[Timothée Lacroix]] (former Meta and Google DeepMind researchers). Valued at over $14B with $3B+ in total funding. Develops both open-weight and proprietary large language models.
## Founders
- **Arthur Mensch** (CEO): Former Google DeepMind
- **Guillaume Lample**: Former Meta AI
- **Timothée Lacroix**: Former Meta AI
## Key models
### Flagship
- **[[Mistral Large 3]]** (Dec 2025): Sparse mixture-of-experts, 41B active / 675B total parameters
- **[[Mistral Medium 3.5]]** (Apr 2026): Open-weights flagship merged model; 128B dense; 256k context; default driver behind Mistral Vibe and [[Le Chat]]
- **[[Mistral Medium]]**: Mid-range line covering versions 3 / 3.1 / 3.5
### Small / Efficient
- **[[Mistral Small 4]]** (Mar 2026): Unified instruct + reasoning + multimodal + coding, MoE 119B total / 6B active, Apache 2.0
- **[[Mistral Small 3.1]]** (Mar 2025): Smaller, efficient model
- **[[Ministral 3]]** (Oct 2024): Small, dense edge/on-device models
### Reasoning
- **[[Magistral Small]]** (Jun 2025): Open-source reasoning model with chain-of-thought
- **[[Magistral Medium]]** (Jun 2025): Proprietary reasoning model
### Speech
- **[[Voxtral Transcribe 2]]** (Feb 2026): State-of-the-art speech-to-text with diarization, 13 languages
### Vision / Multimodal
- **[[Pixtral]]**: Vision-language model family (12B and Large)
### Code
- **[[Codestral]]**: Code generation model
- **[[Devstral]]**: Agentic coding model
## Platform
- **[[La Plateforme]]**: API access and AI Studio
- **[[Le Chat]]**: Consumer chat assistant
- **[[Mistral Compute]]** (planned 2026): European AI infrastructure powered by NVIDIA
## References
- Website: https://mistral.ai
- About: https://mistral.ai/about
- Wikipedia: https://en.wikipedia.org/wiki/Mistral_AI
## Related
- [[Voxtral Transcribe 2]]
- [[Mistral SDK]]
- [[Anthropic]]
- [[OpenAI]]