PeopleofSuperintelligence

Researcher

Szymon Antoniak

Szymon Antoniak is an AI scientist at Mistral AI in Warsaw and an author on papers behind the company's Mixtral, Pixtral and Magistral models. He studied at the University of Warsaw, and in 2023 he coauthored Magnushammer, a transformer-based method for premise selection in automated theorem proving, as a student researcher at IDEAS NCBR. In October 2023, he was first author of Mixture of Tokens, a continuous mixture-of-experts architecture. After joining Mistral AI in January 2024, he took part in Mixtral of Experts, Pixtral 12B in 2024 and Magistral, Mistral's first reasoning model, in 2025. Voxtral, Ministral 3 and Shieldstral followed in 2025 and 2026.

Works
14
Citations
177
h-index
4

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2017–2023

    Student at the University of Warsaw

    He studies at the University of Warsaw, as an undergraduate from 2017 to 2020 and as a master's student from 2021 to 2023.

  2. 2023

    Magnushammer for premise selection

    As a student researcher at IDEAS NCBR in 2023, he coauthors Magnushammer, a transformer-based method for premise selection in automated theorem proving. Magnushammer beats the Sledgehammer tool, reaching 59.5% against 38.3% on the PISA benchmark, and appears at ICLR 2024.

  3. October 2023

    Mixture of Tokens

    In October 2023, he and nine coauthors post Mixture of Tokens, a continuous mixture-of-experts architecture, with him as first author. The best models train 3x faster than dense Transformers in language pretraining and match state-of-the-art MoE architectures.

  4. January 2024

    Joining Mistral AI

    In January 2024, he becomes an AI scientist at Mistral AI in Warsaw. He is among the authors of Mixtral of Experts, whose Mixtral 8x7B model gives each token access to 47B parameters while using 13B active parameters, and of MoE-Mamba, which reaches the performance of Mamba in 2.35 times fewer training steps.

  5. February–October 2024

    Fine-grained MoE and Pixtral

    In February 2024, he coauthors Scaling Laws for Fine-Grained Mixture of Experts, which introduces granularity, a hyperparameter that controls the size of the experts. In October 2024, he is among the authors of Pixtral 12B, a 12-billion-parameter multimodal language model released under Apache 2.0.

  6. June 2025

    Magistral reasoning model

    In June 2025, he is among the authors of Magistral, Mistral's first reasoning model, built with the company's own scalable reinforcement learning pipeline. Mistral open-sources Magistral Small under Apache 2.0.

  7. July 2025–January 2026

    Voxtral and Ministral 3

    In July 2025, he is among the authors of Voxtral, a pair of multimodal audio chat models released under Apache 2.0. In January 2026, he is among the authors of Ministral 3, a family of dense language models in 3B, 8B and 14B sizes.

  8. As of October 2026

    AI scientist at Mistral

    He is an AI scientist at Mistral AI in Warsaw. In July 2026, he is among the authors of Shieldstral, a 3B-parameter multimodal safety classifier.

Career

  1. Since 2024Mistral AIAI scientist
  2. 2023IDEAS NCBRStudent Researcher

Education

  1. 2023University of WarsawMaster's student
  2. 2020University of WarsawUndergraduate student

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

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Szymon Antoniak. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Mistral AI. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Mistral AI. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staffIncludes Mistral AI. 5,000 frontier users would be 38.2% of the core labs' 13,100 reported staff. We test the estimate against 3,913 named model contributors and dated all-staff records across Google DeepMind, OpenAI and Anthropic, plus six additional frontier-grade labs.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Szymon Antoniak (Mistral AI) — Lacuna lacuna.tiptreesystems.com. Checked 20 October 2026.
  2. Magnushammer: A Transformer-Based Approach to Premise Selection (arXiv) arxiv.org. Checked 20 October 2026.
  3. Mixture of Tokens: Continuous MoE through Cross-Example Aggregation (arXiv) arxiv.org. Checked 20 October 2026.
  4. Szymon Antoniak | alphaXiv alphaxiv.org. Checked 20 October 2026.
  5. Szymon Antoniak - AI Scientist at Mistral AI | The Org theorg.com. Checked 20 October 2026.
  6. Mixtral of Experts (arXiv) arxiv.org. Checked 20 October 2026.
  7. MoE-Mamba: Efficient Selective State Space Models with Mixture of Experts (arXiv) arxiv.org. Checked 20 October 2026.
  8. Scaling Laws for Fine-Grained Mixture of Experts (arXiv) arxiv.org. Checked 20 October 2026.
  9. Pixtral 12B (arXiv) arxiv.org. Checked 20 October 2026.
  10. Magistral (arXiv) arxiv.org. Checked 20 October 2026.
  11. Voxtral (arXiv) arxiv.org. Checked 20 October 2026.
  12. Ministral 3 (arXiv) arxiv.org. Checked 20 October 2026.
  13. Shieldstral (arXiv) arxiv.org. Checked 20 October 2026.