Mistral AI announces Mistral Large 4: multimodal LLM with 1T parameters
Mistral AI has unveiled its new flagship model, Mistral Large 4 (codename Le Chonk). This is a large, natively multimodal language model that boasts impressive scale—1 trillion parameters and 49 billion active parameters.
The model is positioned as the best in its class among open-weights models developed in Europe. It demonstrates advanced capabilities in critical areas such as cybersecurity, finance, and industrial manufacturing, surpassing even closed flagship models in visual grounding tasks.
Mistral Large 4 is available via API today, and its European origin and infrastructure (Mistral Cloud) make it an attractive choice for enterprise clients, especially in the cybersecurity sector. The full open-weights release is scheduled for the end of October.
Why it matters
- —Claim of superiority: Mistral states that the model surpasses closed flagship models in key tasks.
- —Geopolitical focus: Creation and deployment of the model entirely in Europe (Mistral Cloud), which is important for the European market.
- —Scale and functionality: The announcement of 1 trillion parameters and native multimodality sets a new standard for open LLMs.
Key facts
- The Mistral Large 4 model has 1 trillion parameters and 49 billion active parameters.
- The model is natively multimodal and positioned as the best open model from Europe.
- Provides an advanced level in cybersecurity, finance, and manufacturing.
- Available via API, as well as for partners in private mode.
- The open weights release is scheduled for the end of October.
The full text is in the original source. Here we provide a brief summary and key facts.