On October 21, 2026, Mistral AI published a post about working with NVIDIA on AI for engineering, aimed at aerospace, automotive and semiconductors. Most of it describes planned work; Mistral gives no dates for the pieces below.
What to know
- Simulation: Mistral says Mistral Large 4, its data generation and physics modeling pipelines, and NVIDIA CUDA-X libraries and NVIDIA Warp are meant to speed up simulation and connect simulation, model development and design.
- 3D: Mistral models are to be integrated with NVIDIA Omniverse libraries for GPU-accelerated 3D scene generation and SimReady asset creation.
- Benchmarks and data: public benchmarks and open datasets for physics models and CAD-to-simulation workflows, plus agentic benchmarks for computer-aided engineering and electronic design automation that measure tool use, multistep workflows, correctness, execution time and compute cost.
- Large 4: Mistral says it is in public preview, about 1 trillion parameters with 49 billion active, trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral-operated European data centers, with weights planned for the end of October 2026.
- Compute: Mistral says its roadmap targets 1 GW of capacity by 2030.
Why it matters
The open question is the weights: until they ship, Large 4 is a hosted preview. See our earlier Large 4 preview report, the Mistral page and open-weight models explained.
Source: Mistral AI. We summarise in our own words; check the source for the full details.