PeopleofSuperintelligence

Researcher

Alexandre Sablayrolles

Alexandre Sablayrolles is a research scientist at Mistral AI. He co-authored Mistral 7B, Mixtral, Magistral and Ministral 3. His education included degrees from École Polytechnique and NYU and a joint CIFRE doctorate with Facebook AI and INRIA, which led to a PhD from Université Grenoble Alpes in 2020. His early research studied membership inference and training-data tracing; he later co-authored DeiT and Opacus, for efficient image transformers and differential privacy.

Works
46
Citations
6,688
h-index
16
Badges

Publication numbers from OpenAlex, as of 18 October 2026.

Timeline

  1. 2017–2020

    A PhD through a CIFRE program

    His education includes an M.S. in Data Science at NYU and B.S. and M.S. degrees in Applied Mathematics and Computer Science at École Polytechnique. He begins a joint CIFRE grant in 2017 under supervision by Cordelia Schmid at INRIA and Hervé Jégou and Matthijs Douze at Facebook AI Research, and receives a PhD from Université Grenoble Alpes in 2020.

  2. 2019–2020

    Membership privacy and data tracing

    In 2019, he and his coauthors derive an optimal strategy for membership inference. In 2020, their Radioactive Data method makes imperceptible changes to image datasets so models can reveal whether a dataset was used in training.

  3. 2020–2021

    Facebook AI, DeiT and Opacus

    By 2021, he is a Research Scientist at Facebook AI in Paris, working on machine-learning privacy and security. He coauthors DeiT, a data-efficient image transformer that uses a teacher-student distillation token, and Opacus, an open-source PyTorch library for differential privacy, in 2020 and 2021, respectively.

  4. 2023–2024

    Mistral 7B and Mixtral

    He coauthors Mistral 7B, a 7-billion-parameter language model, in 2023 and Mixtral 8x7B in 2024, whose layers each contain eight feed-forward experts and route two experts for each token.

  5. 2025

    A reasoning model called Magistral

    He coauthors Magistral, Mistral's first reasoning model; the 2025 paper introduces the team's scalable reinforcement-learning pipeline.

  6. As of October 2026

    Research scientist at Mistral

    As of October 2026, he is a research scientist at Mistral AI. In 2026, he coauthors Ministral 3, a family of parameter-efficient dense language models designed for applications with compute and memory limits.

A language model matched the links in this timeline to people, organizations and papers (provisional, model-judged).

Career

  1. Since 2026Mistral AIResearch scientist
  2. 2020–2021MetaResearch Scientist
  3. 2017–2020Inria
  4. 2017–2020FAIR (Meta)

Education

  1. 2020Université Grenoble AlpesPhD
  2. École PolytechniqueB.S. and M.S.
  3. New York UniversityM.S.

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. Training data-efficient image transformers & distillation through attention2020 · 3,799 citationsWith Hervé Jégou and Matthijs Douze
  2. Going deeper with Image Transformers2021 · IEEE/CVF International Conference on Computer Vision (ICCV) · 1,245 citationsWith Hervé Jégou
  3. Mistral 7B2023 · 322 citationsWith Diego de las Casas and Teven Le Scao
  4. White-box vs Black-box: Bayes Optimal Strategies for Membership Inference2019 · 174 citationsWith Cordelia Schmid, Hervé Jégou and Matthijs Douze
  5. Mixtral of Experts2024 · 141 citationsWith Diego de las Casas and Teven Le Scao

From OpenAlex, as of 18 October 2026.

  1. The median h-index is 55 at universities and 20 at frontier AI labsIncludes Alexandre Sablayrolles. The median h-index is 55 at universities and 20 at frontier AI labs. We compare publishing careers, citations and moves to estimate how universities and frontier labs contribute to AI research.
  2. Google alumni founded or led 79 other organizations, the most among the labs we compareIncludes Mistral AI. We trace 181 organizations founded or led by AI lab alumni, and 187 career moves between the labs.
  3. Google has the highest combined h-index among AI organizationsIncludes Mistral AI. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  4. Of 410 people shaping AI who worked at Google, 161 work there nowIncludes Mistral AI. Our estimates of where former Google and Google DeepMind people work now, including moves within Google.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Alexandre Sablayrolles yet. Send us a free-licensed photo.

  1. IH&MMSec 2021 keynote speakers ihmmsec.org. Checked 18 October 2026.
  2. Project-Team: THOTH radar.inria.fr. Checked 18 October 2026.
  3. White-box vs Black-box: Bayes Optimal Strategies for Membership Inference arxiv.org. Checked 18 October 2026.
  4. Radioactive data: tracing through training arxiv.org. Checked 18 October 2026.
  5. Training data-efficient image transformers & distillation through attention arxiv.org. Checked 18 October 2026.
  6. Opacus: User-Friendly Differential Privacy Library in PyTorch arxiv.org. Checked 18 October 2026.
  7. Mistral 7B arxiv.org. Checked 18 October 2026.
  8. Mixtral of Experts arxiv.org. Checked 18 October 2026.
  9. Magistral arxiv.org. Checked 18 October 2026.
  10. alexandresablayrolles (Alexandre Sablayrolles) · GitHub github.com. Checked 18 October 2026.
  11. Ministral 3 arxiv.org. Checked 18 October 2026.