Faster Results, Clearer Insights: The Promise of Mistral’s Magistral
Mistral AI, one of Europe’s most prominent artificial intelligence companies, has officially released its new line of reasoning models under the name Magistral . The release includes two versions: Magistral Small , an open-source model with 24 billion parameters, and Magistral Medium , a more powerful variant aimed at enterprise clients.
These models are built to excel in complex, domain-specific tasks such as legal analysis, financial modeling, and engineering calculations. One of their standout features is traceable reasoning , which allows users to follow the model’s decision-making process — making it easier to understand, audit, and trust the results. Additionally, Magistral supports multilingual reasoning , performing well on queries in languages like French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese.
🔍 Performance Benchmarks
In internal testing, Mistral compared Magistral against DeepSeek R1 across several benchmarks. On the AIME2024 test, which measures advanced reasoning capabilities:
- Magistral Medium scored 73.6% (up to 90% using majority voting at 64 samples)
- Magistral Small achieved 70.7% (83.3% with majority voting)
Like DeepSeek R1, Magistral now offers step-by-step reasoning output , allowing users to see how it arrives at conclusions. It also introduces a new Flash Answers mode , which delivers responses up to 10 times faster than competing models — a major advantage for time-sensitive applications.
📈 Real-World Applications
Thanks to its specialized training, Magistral is designed to support industries that require high accuracy and accountability. This makes it particularly useful for:
- Legal research : Analyzing case law and generating summaries
- Financial forecasting : Supporting data-driven investment decisions
- Regulated sectors : Providing transparent outputs for compliance and audits
The ability to trace each step of the model’s logic is especially valuable in regulated environments where explainability is crucial.
📦 Availability and Access
Mistral is making Magistral accessible through multiple channels:
- Magistral Small is available under the Apache 2.0 license on Hugging Face , encouraging community use and development.
- Magistral Medium can be accessed via Mistral’s Le Chat platform or through an API on La Plateforme .
- The model will soon be available on leading cloud platforms including:
- Amazon SageMaker
- IBM WatsonX
- Azure AI
- Google Cloud Marketplace
For individual users looking to try it out, the easiest way is to visit chat.mistral.ai , where they can interact with Magistral directly through the Le Chat interface .
