What was announced
Mistral AI said it is working with NVIDIA to build a new generation of AI-native engineering workflows. The goal is to speed up design and simulation cycles.
What is being used
- The Mistral Large 4 model
- Mistral's data generation and physics modeling pipelines
- NVIDIA CUDA-X libraries and NVIDIA Warp
- NVIDIA Omniverse libraries for GPU-accelerated 3D scene generation and SimReady assets
The idea is to connect simulation, AI model development and engineering design in a closed loop.
Open benchmarks in focus
A central part of the collaboration is public benchmarks and open datasets for evaluating physics models. Neil Ashton, VP of Computational Engineering at Mistral, said these will give researchers and industrial teams a shared foundation to compare accuracy and validate automated CAD-to-simulation workflows.
Target industries
Aerospace, automotive and semiconductors are the main targets. The two companies will also develop evaluations for agentic AI in computer-aided engineering (CAE) and electronic design automation (EDA), measuring whether agents can use engineering tools, complete multistep workflows and produce technically correct results.
