# Arthur Mensch

> Arthur Mensch is the French machine-learning researcher who left Google DeepMind in 2023 to co-found and lead Mistral AI. With longtime peers Guillaume Lample and Timothée Lacroix, he built a Paris company around efficient models, portable deployment and user control, then expanded from open-weight releases into products, custom model tooling and European compute infrastructure. By September 2026, Mistral said it operated in 20 countries, supported more than 125 global enterprises and had raised €3 billion at a post-money valuation above €21 billion.

- **Role:** Co-founder and CEO at Mistral AI
- **Organizations:** Mistral AI, Google DeepMind, École normale supérieure, Inria, NeuroSpin
- **From:** Sèvres, France
- **Nationality:** French
- **Education:** Engineering studies, class of 2011, École Polytechnique, Engineering studies, Télécom Paris, Master MVA (Mathematics, Vision, Learning), Université Paris-Saclay / ENS Paris-Saclay, PhD in machine learning and functional MRI, Université Paris-Saclay, Inria and NeuroSpin
- **Known for:** Co-founded Mistral AI and shipped its first open-weight model within months of launch., Co-authored DeepMind’s influential 2022 paper Training Compute-Optimal Large Language Models, which introduced Chinchilla., Co-authored research applying machine learning across large collections of functional MRI studies.

## Career timeline

- **2015-2018** — Completed doctoral research at Inria and NeuroSpin on machine learning, stochastic optimization and representations learned from functional MRI data.
- **2018-2020** — Worked as a postdoctoral researcher at École normale supérieure on optimal transport and stochastic optimization, with research time at NYU’s Courant Institute.
- **2020-2023** — Worked at Google DeepMind on large language models and co-authored the 2022 Chinchilla compute-scaling paper.
- **2023-04** — Co-founded Mistral AI in Paris with Guillaume Lample and Timothée Lacroix.
- **2023-09** — Mistral released Mistral 7B, its first open-weight model, four months after hiring its first employee.
- **2024** — Mistral expanded its model range and distribution; Mensch was named to both TIME100 AI and TIME100 Next.
- **2025** — Mistral launched Le Chat, OCR, coding and speech products, plus Mistral Compute; a €1.7 billion round led by ASML reinforced its infrastructure strategy.
- **2026-05** — At Mistral’s AI Now Summit, Mensch set out a full-stack strategy spanning chips, data centers, models and applications.
- **2026-09** — Mistral announced a €3 billion Series D led by Samsung Electronics at a post-money valuation above €21 billion.

## Achievements

- Co-founded Mistral AI and shipped its first open-weight model within months of launch.
- Co-authored DeepMind’s influential 2022 paper Training Compute-Optimal Large Language Models, which introduced Chinchilla.
- Co-authored research applying machine learning across large collections of functional MRI studies.
- Named to TIME100 AI 2024 and TIME100 Next 2024.
- Led Mistral’s expansion from foundation models into applications, custom model tooling and European compute infrastructure.
- Led the company through a September 2026 Series D that Mistral said was Europe’s largest technology equity fundraising round.

## Latest updates

- **2026-09** — Mistral announced a €3 billion Series D led by Samsung Electronics at a post-money valuation above €21 billion; it said it operated in 20 countries and supported more than 125 global enterprises.
- **2026-09** — Mistral and Mozilla announced that Mistral models would power Firefox Smart Window in France and North America, with the UK and Germany expected to follow.
- **2026-08** — Mistral expanded its regional inference and sovereign AI infrastructure strategy, including support for third-party open models.
- **2026-05** — Mensch presented Mistral’s full-stack direction at the company’s AI Now Summit, including additional European data-center capacity and industrial partnerships.
- **2026-01** — At Axios House in Davos, Mensch argued that enterprise AI adoption contains more internal friction than public claims suggest and works best when companies reorganize specific functions.

## Links

- Website: https://amens.ch
- LinkedIn: https://www.linkedin.com/in/arthur-mensch
- Twitter/X: https://x.com/arthurmensch
- GitHub: https://github.com/arthurmensch
- YouTube: https://www.youtube.com/watch?v=xgaLsQTFUEw

---

Profile page: https://yespress.io/arthur-mensch
Published by YesPress — https://yespress.io
Last updated: 2026-09-26