Mistral and Nvidia have announced a partnership to accelerate engineering simulations and industrial product design using artificial intelligence. The collaboration combines Mistral Large 4 with the company’s physics AI alongside Nvidia’s CUDA-X, Warp software libraries, and Omniverse platform for three-dimensional simulations. This integration aims to connect design, simulation, and AI training into a single continuous workflow for sectors including aerospace, automotive, and semiconductors.
The partners intend to release public benchmarks and open datasets to allow independent testing and comparison of physics models and AI agents within engineering software. This follows their previous work on the Mistral NeMo model and current efforts within the Nemotron Coalition to build an open foundation model. Mistral Large 4 contains approximately one trillion parameters and was trained on 3,800 Grace Blackwell GPUs located in Europe. The model is scheduled to ship as an open-weight version in late October.
* Mistral targets an expansion of computing capacity to one gigawatt by 2030.
* The new workflow links design directly with simulation and AI training.
* Benchmarks will focus on physics models and AI agents for engineering tasks.
