Magistral Medium (latest)

Magistral Medium is Mistral AI's first dedicated reasoning model, positioned as the larger sibling in the Magistral family and offered as an API-only deployment rather than as open weights. The model was announced on 10 June 2025 alongside a smaller open-weights variant, marking Mistral's entry into the reasoning-focused model category alongside competitors in the space. By concentrating compute behind an API surface, the Medium tier is intended for users who need stronger reasoning behavior without self-hosting the larger weights.

Practically, Magistral Medium fits workflows that benefit from explicit step-by-step inference, such as multi-step problem solving, code reasoning, and analytical tasks where traceable logic matters. The family is built around Mistral's reasoning approach, with the Medium tier offering higher capacity than the 24B Apache-2-licensed open-weights Magistral Small, while still being accessible through a managed endpoint. For teams already using Mistral models, the Magistral Medium tier adds a reasoning-oriented option that complements rather than replaces general-purpose chat models, making it well suited as a specialized component in pipelines that require deliberation rather than open-ended generation.