MistralReleased on Jun 10, 2025

Magistral Medium: API Pricing, Context Window & Benchmarks

Magistral Medium is a language model from Mistral, released in June 2025, with multimodal input.

Trained solely with reinforcement learning on top of Mistral Medium 3, Magistral Medium is a reasoning model that achieves strong performance on complex math and code tasks without relying on distillation from existing reasoning models.

Magistral Medium benchmarks

Rankings

Quality Tracker

Magistral Medium Performance Across Datasets

Scores sourced from the model's scorecard, paper, or official blog posts

LLM Stats Logollm-stats.com - Sat Jul 18 2026
Notice missing or incorrect data?

Magistral Medium model size

Magistral Medium has 24 billion parameters. See how it compares to other models in the same parameter range.

Parameters
24B
Medium (10–30B)
24B
1B7B70B405B

Magistral Medium API

Available from the model provider

Magistral Medium has an official provider API. It is not currently routed through the LLM Stats gateway.

Read the official API documentation

Magistral Medium latency

Magistral Medium time to first token, sustained output throughput, and failed-request rate from live API traffic over the trailing 7 days.

Magistral Medium examples

Recent arena outputs from Magistral Medium, picked from the highest-ranked matchups.

Magistral Medium license

Magistral Medium is released under the Apache 2.0 license, which permits commercial use, has 24.0B parameters, has a knowledge cutoff of June 2025.

License
Apache 2.0
Commercial use allowed
Parameters
24.0B
Knowledge cutoff
June 2025

Apache License 2.0 - allows commercial use

Magistral Medium resources

Official sources for Magistral Medium: api documentation, official playground, paper or system card, official launch post.

Magistral Medium vs other models

The most-compared alternatives to Magistral Medium are DeepSeek R1 Zero, Phi 4 Reasoning, DeepSeek R1 Distill Qwen 14B. Open any pair side-by-side for benchmarks, pricing, context, and latency.

Models like Magistral Medium

Models ranked just above and below Magistral Medium by LLM Stats score.

 

DeepSeek R1 Zero

Score pending
DetailsCompare
 

Phi 4 Reasoning

Score pending
DetailsCompare
 

DeepSeek R1 Distill Qwen 14B

Score pending
DetailsCompare
 

Qwen3 VL 32B Instruct

Score pending
DetailsCompare
 

LongCat-Flash-Lite

Score pending
DetailsCompare
 

Qwen3 VL 30B A3B Instruct

Score pending
DetailsCompare

FAQ

Common questions about Magistral Medium.

When was Magistral Medium released?

Magistral Medium was released on June 10, 2025 by Mistral. This is the official Magistral Medium release date tracked on LLM Stats.

Is Magistral Medium available via API?

Yes, Magistral Medium is available via API. See the official documentation for authentication and endpoint details.

How big is Magistral Medium?

Magistral Medium has 24 billion parameters. It ships as an open-weight model, so you can download and run it on your own hardware.

Who created Magistral Medium?

Magistral Medium was created by Mistral.

What is the license for Magistral Medium?

Magistral Medium is released under the Apache 2.0 license. This is an open-source / open-weight license that permits self-hosting.

What is the knowledge cutoff date for Magistral Medium?

Magistral Medium has a knowledge cutoff of June 2025, meaning it was trained on data up to that point and may not know about events after it.

Is Magistral Medium multimodal?

Yes, Magistral Medium is multimodal and can accept both text and images as input.

Where is the Magistral Medium paper or technical report?

Magistral Medium has a paper or technical report available at https://arxiv.org/pdf/2506.10910. Use that source for architecture, training, release and evaluation details.

What models should I compare Magistral Medium against?

Common Magistral Medium comparisons include Magistral Medium vs DeepSeek R1 Zero, Magistral Medium vs Phi 4 Reasoning, Magistral Medium vs DeepSeek R1 Distill Qwen 14B. Compare them side by side for benchmark scores, pricing, context window, latency and API availability.