Mistral AI joins NVIDIA Nemotron Coalition to accelerate open frontier model development

Mistral AI has announced its joining as a co-founder of the NVIDIA Nemotron Coalition, a global initiative bringing together leading AI labs to advance open frontier models. As part of a strategic partnership, Mistral AI and NVIDIA plan to jointly develop open-source AI models, combining Mistral AI's specialized architecture and platform with NVIDIA's computing resources, model development tools, and synthetic data generation pipelines. Mistral AI will contribute its own training methods, multimodal capabilities, and fine-tuning tools, leveraging NVIDIA's infrastructure for scaling. This continues previous scientific collaboration, including the development of the Mistral Nemo model. As part of its commitment to open models, the company has also released Mistral Small 4. The coalition's first initiative will be a base model trained on NVIDIA DGX Cloud, which will serve as the foundation for the NVIDIA Nemotron 4 family — these models will be published as open source. Mistral AI brings to the coalition large-scale model development, multimodal capabilities, modern architectures, and expertise in customization through pretraining, post-training, and fine-tuning.