Lélio Renard Lavaud

Also known as Lélio Renard-Lavaud

Lélio Renard Lavaud is VP of Engineering at Mistral AI and a co-author of the technical reports behind several of its language models. They co-authored Mistral 7B in October 2023 and Mixtral of Experts in January 2024, two models released under the Apache 2.0 license. In 2025, they co-authored the reports on Magistral, Mistral's first reasoning model, the Voxtral audio chat models and Devstral-Small, a model for code agents. At AI Engineer Paris in September 2025, they argued that adoption at enterprise scale hinges on openness. In September 2026, they returned to the event at Station F with a main-stage talk titled "Building Frontier AI".

Now at: Mistral AI

Timeline

October 2023

Co-authoring Mistral 7B

In October 2023, they co-author Mistral 7B, a 7-billion-parameter language model that uses grouped-query and sliding window attention and outperforms Llama 2 13B on every evaluated benchmark. The base model and its instruction-tuned version are released under the Apache 2.0 license.
January 2024

Mixtral of Experts

They co-author Mixtral of Experts in January 2024, describing Mixtral 8x7B, a sparse mixture-of-experts model in which a router picks two of eight feedforward blocks for each token at every layer. Its base and instruct versions are released under the Apache 2.0 license.
June 2025

Magistral, a reasoning model

In June 2025, they co-author the Magistral report, which introduces Mistral's first reasoning model and the company's own reinforcement learning pipeline. Magistral Medium is trained with reinforcement learning alone on top of Mistral Medium 3, and Magistral Small is open-sourced under Apache 2.0.
July–August 2025

Voxtral and Devstral

In July 2025, they co-author the Voxtral report on Voxtral Mini and Voxtral Small, two multimodal audio chat models released under Apache 2.0. Their name also appears on the Devstral report of August 2025, which presents Devstral-Small, a 24B open source model for code agents.
September 2025

Openness at AI Engineer Paris

Speaking for Mistral at AI Engineer Paris in September 2025, they argue that adoption at enterprise scale hinges on openness, with open models sidestepping vendor lock-in and adding transparency.
2025–2026

A Kong Konnect customer

In November 2025, Kong invites them to an executive dinner in Paris as Mistral AI's VP of Engineering and a Kong Konnect customer, to share how open AI models go into production at scale. Kong's 2026 live stream on its AI Connectivity roadmap features them saying that Mistral AI's mission is to put AI in everyone's hands.
June 2026

Off the Radar in Brussels

At Off the Radar in Brussels in June 2026, a private gathering of builders, CTOs and engineers, they share a session as Mistral's VP of Engineering with Romain Dillet, founder of Hypertext.
As of October 2026

VP of Engineering at Mistral

They are VP of Engineering at Mistral AI, which brings the AI Engineer conference back to Paris in September 2026. Their talk "Building Frontier AI" runs on the main stage at Station F during the September 2026 event.