Mistral AI launched its new 1 trillion-parameter multimodal AI model, Mistral Large 4 (ML4), in Europe on October 6, 2026, challenging existing models with its advanced capabilities.
The model, nicknamed “le Chonk,” is available as a public preview through an API on Mistral Studio, marking a development in the European artificial intelligence sector.
Mistral AI stated that ML4 was trained using 3,800 NVIDIA GPUs located in the company’s European datacenters. The training data incorporates more than 160 languages, including all official languages of the European Union.
The company asserted that ML4 performs comparably to leading open-source models globally and surpasses all open models developed in the United States or Europe in key areas such as code generation, agent tasks, and cybersecurity.
The weights for ML4 are scheduled for publication at the end of October, which will allow companies to self-host and independently utilize the model, according to Mistral AI.
ML4 has demonstrated strong performance in cybersecurity applications, code generation, and agent-based tasks. It also features vision capabilities, positioning it as a comprehensive, European-operated, end-to-end solution.
The introduction of ML4 emphasizes European data sovereignty and aims to provide a competitive alternative to models primarily developed outside the continent, according to statements from Mistral AI.
Analysts at Artificial Analysis, Surge AI, and vals.ai have previously evaluated various AI models, noting the increasing competition in the field. Mistral AI’s latest offering directly enters this competitive environment, particularly against models from organizations like OpenAI.
Source: Abondance
