French AI lab Mistral AI unveiled Mistral Large 4, nicknamed Le Chonk, a 1 trillion-parameter multimodal model on October 6. The model features 49 billion active parameters and is available in public preview via Mistral Studio API, with the full open-weight release planned by the end of the month, according to mistral.ai.
Le Chonk aims to surpass both closed and open-source rivals by delivering frontier performance in coding, agentic workflows, and multimodal understanding. Mistral is currently conducting safety testing and red-teaming with cybersecurity leaders, vetted partners, and state authorities who access the model with reduced moderation. Pierre Stock, Mistral’s VP of Science, said the model was trained entirely on Mistral’s own compute infrastructure, using 4,000 NVIDIA GPUs—two to three times fewer than Chinese competitors and significantly less than closed-source models, as reported by techcrunch.com.
The model’s performance is competitive with the strongest open-source models globally and outperforms any open-weight model developed in the US or Europe. It demonstrates state-of-the-art results in enterprise workloads such as cybersecurity, finance, and law, and exceeds frontier closed models in visual grounding tasks. Mistral positions ML4 as a European alternative amid a growing divide between closed models and open models primarily developed in China, aligning with French President Macron’s vision of a “third way in AI.”
Mistral plans to release the full open-weight model weights by the end of October after completing safety evaluations. Until then, the model remains accessible via a guarded endpoint, with trusted partners and governments collaborating to ensure its use for defense rather than malicious purposes, Pierre Stock confirmed to techcrunch.com.