NVIDIA opens four Cosmos 3 world models for driving and robot simulation

OFFICIAL Written by GPN · Nov 6, 2026, 6:40 PM ET · NVIDIA Cosmos 3 · Tech

NVIDIA's Cosmos world-model platform has four new open-weight checkpoints. The Cosmos 3 additions cover interactive world simulation and multi-sensor driving scenes, each post-trained from Cosmos 3 Nano with 16 billion trainable parameters and released under the OpenMDW1.1 licence, which permits commercial and non-commercial use.

Cosmos3-Nano-Transfer-Auto generates multi-view driving data with varied weather, lighting and time of day, producing synchronised video across up to 11 camera views, including fisheyes, with joint RGB and LiDAR generation. Its distilled sibling, Cosmos3-Nano-Sim-Auto, streams synchronised camera and LiDAR observations in small chunks for fast, interactive rollouts.

Cosmos3-Nano-Sim-Transfer turns a physics simulator's controls and a text prompt into photorealistic video, using depth, segmentation, edge and blur controls while the simulator keeps the ground-truth physics. Cosmos3-Nano-Sim-Bimanual predicts video from a two-arm robot's actions, supporting interactive simulation for YAM and AgiBot embodiments.

NVIDIA's aim is an open foundation for building richer simulated worlds: open weights and post-training recipes give researchers a shared starting point for adapting the models, comparing methods and contributing improvements. Checkpoints, a technical report and demos are available now.

Facts gathered from

Hugging Face (NVIDIA)

NVIDIA opens four Cosmos 3 world models for driving and robot simulation

Facts gathered from

More on this