NVIDIA Cosmos 3 — first open omni-model for physical AI reasoning and action

Summary

NVIDIA released Cosmos 3, an open-source omni-model for physical AI that unifies world generation, physical reasoning, and action generation in a single model. It replaces the prior split architecture (separate Cosmos Predict, Transfer, Reason, and Policy models) with one Mixture-of-Transformers backbone that processes text, image, video, audio, and action in a single forward pass. The design uses separate autoregressive and diffusion subsequences sharing joint attention, letting one model act a