Mistral Large 4: 1,000 Billion Parameters and 38 Points at Artificial Analysis

Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Mistral launches Mistral Large 4, distributed via its API and announced in open weights starting in October. On the Artificial Analysis index, the model jumps from 9 points for the previous generation to 38, placing France in third position among "frontier" AI countries, behind the United States and China.
38 points on the Artificial Analysis index and a country rank of #3
Artificial Analysis assigns 38 points to Mistral Large 4 on its intelligence index, compared to 9 points for Mistral Large 3. According to this source, Mistral Large 4 would have performance equivalent to OpenAI's GPT-6 Luna and would reach the same level as GLM-5.3-Flash on the cyber index. Artificial Analysis also positions the model close to DeepSeek V4.1, GLM-5.3-Flash, and Kimi K3. Based on this ranking, France would hold the third place by country in "frontier" AI, behind the United States (Claude Opus 5.5 (max), 58) and China (MiMo-V2.6-Pro, 46).
Model available via API and open weights publication announced for October
Mistral Large 4 is accessible via Mistral's API, and the publication of its open weights is scheduled for October. This model boasts 1 trillion parameters. Mistral has officially announced this launch and has dubbed it "Le Chonk."
Targeted use cases, claimed visual anchoring, and promise of sovereignty
According to Mistral, the model targets use cases in cybersecurity, industrial applications, and finance, and outperforms closed AIs in visual anchoring. It is suggested that this offering would allow France and the European Union to limit performance concessions when a sovereign option is sought, and would strengthen Mistral's positioning against American and Chinese models. Mistral Large 4 is presented as one of the best open models, a category dominated by Chinese laboratories, which companies can adapt and deploy on their own servers. This model represents a significant evolution compared to the previous version.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.