Europe's top AI lab wants its models to do the simulation work behind cars, planes and chips. Mistral and Nvidia announced on Wednesday, Oct. 21, 2026, at Nvidia's GTC Berlin conference, that they will build "AI-native engineering workflows" combining Mistral Large 4 with Nvidia's CUDA-X libraries, the Warp physics framework and Omniverse 3D tools, Mistral said in a blog post.

What they're building

  • Faster simulation: Mistral's data-generation and physics-modeling pipelines, accelerated by CUDA-X and Warp, aimed at closed-loop workflows that connect simulation, AI model training and design.
  • 3D assets: integrating Mistral's models with Omniverse libraries for GPU-accelerated 3D scene generation and "SimReady" asset creation.
  • Public benchmarks: open datasets and benchmarks for physics models, plus evaluations of AI agents in computer-aided engineering (CAE) and chip design (EDA) that measure tool use, multistep workflows, correctness, run time and compute cost.

Mistral named aerospace, automotive and semiconductors as target industries. Nvidia's blog said Mistral plans to use CUDA-X, Warp and Omniverse to accelerate physics simulation and generate training data for engineering AI models. Neither company gave dates for the benchmarks or datasets.

Mistral Large 4

The Decoder reported that Mistral Large 4 has about one trillion parameters, was trained on 3,800 Grace Blackwell GPUs in Europe, and is set to ship as an open-weight model in late October. Mistral wants to expand its computing capacity to one gigawatt by 2030, The Decoder said.

Why it matters for Tesla watchers

The Berlin event also produced Nvidia's European self-driving deals, including Bolt's Lucid robotaxis on Alpamayo and VW and Porsche's shared driving-data loop, as European carmakers lean on Nvidia's stack while Tesla builds its own chips and software.