{"id":25316,"date":"2025-06-11T17:16:11","date_gmt":"2025-06-11T15:16:11","guid":{"rendered":"https:\/\/comunicagenia.com\/blog\/mistral-ai-masterclass-the-first-european-reasoning-model\/"},"modified":"2025-06-11T17:16:11","modified_gmt":"2025-06-11T15:16:11","slug":"mistral-ai-masterclass-the-first-european-reasoning-model","status":"publish","type":"post","link":"https:\/\/comunicagenia.com\/en\/blog\/mistral-ai-masterclass-the-first-european-reasoning-model\/","title":{"rendered":"Mistral AI Masterclass: The first European reasoning model"},"content":{"rendered":"

Last Tuesday the French startup Mistral AI<\/span><\/a> took a significant step in the field of artificial intelligence with the launch of <\/span>Magistral, its first family of reasoning models.\u00a0<\/b><\/p>\n

Composed of two variants (Magistral Small and Magistral Medium) this line of models is designed to address complex tasks, from structured calculations to programmatic logic and decision trees.\u00a0<\/span><\/p>\n

Thanks to its focus on traceability and transparency, Magistral is positioned as an especially valuable option for regulated sectors such as legal, financial and healthcare.<\/span><\/p>\n

The Mistral AI approach<\/span><\/h2>\n

Since its founding in 2023, Mistral AI has made it its mission to offer powerful, accessible and efficient AI solutions.<\/span><\/p>\n

With Magistral, the company reinforces that vision by presenting <\/span>an open model and another oriented to companies<\/b>, which reinforces its competitiveness against giants such as OpenAI or DeepSeek.\u00a0<\/span><\/p>\n

At the same time, the commitment to a European origin and the native support of more than twenty languages reinforce its ambition as a solid alternative in the global market.<\/span><\/p>\n

Characteristics of Magistral Small<\/span><\/h2>\n

Magistral Small is a 24 billion parameter model released under the Apache 2.0 license.\u00a0<\/span><\/p>\n

This <\/span>allows any developer to download the model from Hugging Face and deploy it locally<\/b>, even on modest machines like a 32GB MacBook or a PC with a single RTX 4090 GPU.<\/span><\/p>\n

Its context window reaches 128,000 tokens, although Mistral recommends limiting sessions to around 40,000 tokens to maintain optimal performance.<\/span><\/p>\n

Unlike other models, <\/span>Magistral Small integrates a thought chain reasoning mechanism<\/b> delimited by `<think> \u2026 <\/think>`, which documents each logical step before offering the final answer.<\/span><\/p>\n

This makes it easier to audit and understand your internal processes. The model does not yet support direct function calls, but it compensates for that limitation with its <\/span>ability to decompose problems into multiple steps<\/b> and generate clear and structured final summaries.<\/span><\/p>\n

Its open source nature<\/b> makes Small an ideal tool for researchers and academic environments, where adaptability and collaborative experimentation are essential.<\/span><\/p>\n

Magistral Medium: performance and speed for companies<\/span><\/h2>\n

Magistral Medium is the variant designed for business environments. <\/span>It is offered in preview through Le Chat and the Mistral API<\/b>, with imminent deployment on platforms such as Amazon SageMaker, IBM Watson X, Azure AI and Google Cloud Marketplace.<\/span><\/p>\n

Its technical innovations include two notable modes: \u201c<\/span>Pure Thinking<\/span><\/i>, optimized for mathematical and coding tasks, and \u201c<\/span>Speed 10x<\/span><\/i>\u201d, based on<\/span> Cerebras<\/b> acceleration technology, which multiplies the performance per token by ten compared to competing models.<\/span><\/p>\n

In addition, Magistral Medium uses a sophisticated <\/span>training<\/span><\/a> scheme. After starting from a Mistral Medium 3 checkpoint, reinforcement learning with verifiable rewards (RLVR) was applied, achieving almost a 50% improvement in the AIME-24 benchmark for the pass@1 metric.<\/span><\/p>\n

This training was complemented by <\/span>SFT (supervised fine-tuning)<\/b> refinement with reasoning traces, which further enhances its reliability in critical applications.<\/span><\/p>\n

Benchmark results and comparisons<\/span><\/h2>\n

In the AIME-2024 evaluation, Magistral Medium achieved a <\/span>73.6% success rate in pass@1<\/b>, rising to 90% with multiple voting, while Magistral Small obtained 70.7% and 83.3% respectively.\u00a0<\/span><\/p>\n

In other tests, such as GPQA Diamond or LiveCodeBench, models from OpenAI and Google, for example <\/span>Gemini 2.5 Pro or Claude Opus 4, slightly outperform Magistral<\/b>.<\/span><\/p>\n

However, the differential value of the Magistral family lies in the traceability of its chains of thought and its notable efficiency, which in the case of the Small version allows local executions without the need for large-scale infrastructures.<\/span><\/p>\n

These results demonstrate that, although Mistral AI does not aspire to lead all raw performance rankings, <\/span>its strategy focuses on offering balanced models<\/b> that combine speed, transparency and multilingualism.\u00a0<\/span><\/p>\n

This makes them especially attractive in contexts where interpretability and control of each step of reasoning is more important than mere raw power.<\/span><\/p>\n

In what areas can Magistral be applied?<\/span><\/h2>\n

Magistral models adapt to multiple professional scenarios. In the legal sector, they facilitate the analysis of contracts and judgments by showing each step of reasoning, reducing the risk of bias and errors.\u00a0<\/span><\/p>\n

In finance, they are used to model risks, generate predictions and audit complex calculations. In healthcare, they support clinical decisions through detailed and traceable explanations.\u00a0<\/span><\/p>\n

Likewise, in software development and project planning, its chain of thought structure enhances the generation of robust and transparent programmatic logic.<\/span><\/p>\n

Another highlight is its multilingual capacity. Magistral supports more than twenty languages, including English, French, Spanish, German, Italian, Arabic, Russian, and Simplified Chinese.\u00a0<\/span><\/p>\n

This makes it <\/span>a suitable option for global companies and international projects <\/b>that require consistency and quality in different languages.<\/span><\/p>\n

Reception and future prospects<\/span><\/h2>\n

The AI community has welcomed Magistral with great interest. On platforms like\u00a0<\/span><\/p>\n

However, temporary shortcomings have been pointed out, such as the absence of function calls in the Small version or certain API restrictions in Medium.<\/span><\/p>\n

Under the direction of Arthur Mensch, Mistral AI aims to consolidate Magistral as a European reference in inference models with deep reasoning.<\/span><\/p>\n

In the medium term,<\/span> cloud integrations are expected to spread<\/b> and extensions for function calls and other input formats to emerge. In parallel, the open source of the Small model will enable community innovations that could further improve its performance and usability.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

Last Tuesday the French startup Mistral AI took a significant step in the field of artificial intelligence with the launch of Magistral, its first family of reasoning models.\u00a0 Composed of two variants (Magistral Small and Magistral Medium) this line of models is designed to address complex tasks, from structured calculations to programmatic logic and decision […]<\/p>\n","protected":false},"author":9,"featured_media":18951,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_uag_custom_page_level_css":"","_thinkrank_schema_form_data":"","_thinkrank_selected_schema_type":"","_thinkrank_additional_schemas":"","_thinkrank_canonical_url":"","_thinkrank_og_title":"","_thinkrank_og_description":"","_thinkrank_og_image":"","_thinkrank_twitter_title":"","_thinkrank_twitter_description":"","_thinkrank_twitter_image":"","_thinkrank_imported_from":"","_thinkrank_focus_keywords":[],"_thinkrank_focus_keyword":"","_thinkrank_robots_meta_enabled":0,"_thinkrank_robots_meta":"","_thinkrank_advanced_robots_meta":"","_thinkrank_exclude_from_search":0,"_thinkrank_exclude_from_archives":0,"_thinkrank_primary_category":0,"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[719],"tags":[],"class_list":["post-25316","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-artificial-intelligence"],"uagb_featured_image_src":{"full":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento.jpg",1408,768,false],"thumbnail":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento-150x150.jpg",150,150,true],"medium":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento-300x164.jpg",300,164,true],"medium_large":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento-768x419.jpg",768,419,true],"large":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento-1024x559.jpg",1024,559,true],"1536x1536":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento.jpg",1408,768,false],"2048x2048":["https:\/\/comunicagenia.com\/wp-content\/uploads\/2025\/06\/magistral-mistral-ai-modelo-europeo-razonamiento.jpg",1408,768,false]},"uagb_author_info":{"display_name":"marketing","author_link":"https:\/\/comunicagenia.com\/en\/blog\/author\/marketing\/"},"uagb_comment_info":0,"uagb_excerpt":"Last Tuesday the French startup Mistral AI took a significant step in the field of artificial intelligence with the launch of Magistral, its first family of reasoning models.\u00a0 Composed of two variants (Magistral Small and Magistral Medium) this line of models is designed to address complex tasks, from structured calculations to programmatic logic and decision…","_links":{"self":[{"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/posts\/25316","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/comments?post=25316"}],"version-history":[{"count":0,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/posts\/25316\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/media\/18951"}],"wp:attachment":[{"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/media?parent=25316"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/categories?post=25316"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/comunicagenia.com\/en\/wp-json\/wp\/v2\/tags?post=25316"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}