PARTNERSHIP · COMPANIES · #2361
Mistral and Nvidia pair Mistral Large 4 with CUDA-X and Omniverse for physics AI
Mistral and Nvidia announced a partnership to combine Mistral Large 4 and Mistral's physics AI with Nvidia's CUDA‑X, Warp libraries and Omniverse to create an end‑to‑end workflow linking design, 3D simulation and AI training for industries like aerospace, automotive and semiconductors. They also plan to publish public benchmarks and open datasets; Mistral Large 4 is reported at about one trillion parameters, trained on 3,800 Grace Blackwell GPUs and due as an open‑weight model in late October.
KEY POINTS
- Mistral and Nvidia announced a partnership to combine Mistral Large 4 and Mistral's physics AI with Nvidia's CUDA‑X, Warp libraries and Omniverse to create an end‑to‑end workflow linking design, 3D simulation and AI training for industries like aerospace, automotive and semiconductors.
- They also plan to publish public benchmarks and open datasets; Mistral Large 4 is reported at about one trillion parameters, trained on 3,800 Grace Blackwell GPUs and due as an open‑weight model in late October.
- Linking an open large model and physics AI with Nvidia's simulation and GPU software could materially speed engineering simulations and set shared benchmarks/datasets for industry adoption.
WHY IT MATTERS
Linking an open large model and physics AI with Nvidia's simulation and GPU software could materially speed engineering simulations and set shared benchmarks/datasets for industry adoption.