Top AI Product

NVIDIA Cosmos Reason 2 is an open reasoning vision-language model with a narrow but hard job: let machines see a physical scene, understand what’s happening, and decide how to act. It’s the reasoning brain in NVIDIA’s broader Cosmos physical-AI stack.

## Reasoning, not just captioning

Most vision-language models describe an image. Cosmos Reason 2 is tuned to reason about it — spatial relationships, physics, cause and effect, what a robot or vehicle should do next. That’s the gap between “there is a cup on the table” and “the cup is near the edge, so approach from the left.” It feeds downstream systems like NVIDIA’s GR00T humanoid models, which call Cosmos Reason for contextual understanding.

## Why open matters here

Releasing it with open weights lets robotics and autonomous-vehicle teams build on a shared reasoning layer instead of each training one from scratch. Paired with Cosmos world models and action models, it rounds out a stack covering perception, reasoning, and action — the three things any physical-AI system has to do, now available as separable open pieces.