Mistral and NVIDIA are collaborating to build AI-native engineering workflows. They will combine Mistral Large 4 with NVIDIA CUDA-X and Warp for faster simulations. The partnership integrates Mistral models with NVIDIA Omniverse for 3D scene generation. Public benchmarks and open datasets will standardize physics and agentic AI evaluations. These will assess model accuracy and performance in engineering tasks. The initiative targets aerospace, automotive and semiconductor industries. Mistral and NVIDIA will define standards for agentic and physics AI evaluations. The work builds on prior collaborations like Mistral NeMo and the Nemotron Coalition. Mistral Large 4 was trained on 3,800 NVIDIA GPUs in European data centers. Mistral plans to release its weights at the end of October. The companies aim to expand AI infrastructure to 1 GW of capacity by 2030.
What changed
Engineering simulations and AI model development were separate, slower workflows without shared benchmarks.
What it unlocks
Faster simulation cycles and closed-loop AI-driven design workflows for engineering teams.
- 1 trillion total parameters
- 49 billion active parameters
- 3,800 NVIDIA Grace Blackwell GPUs
- 1 GW capacity target by 2030
What you need to act on it
- NVIDIA accelerated computing stack
- Mistral AI models
Sources