Cosmos Reason 2
Based on Qwen3-VL (8B variant from Qwen3-VL-8B-Instruct, 8.7B params, 32 GB+ GPU). Initial release 2025-12-19, updated 2026-03-10; promoted at CES 2026. Up to 256K input tokens. Its role is folded into Cosmos 3 for new projects.
- Context window
- 256,000 tokens
- Input
- text, image, video
- Output
- text
- License
- NVIDIA Open Model License (commercial use allowed; HF repo gated - accept terms)
- Verified
- 2026-09-29
How to call it
| Provider | Model id | Endpoint / URL | Docs |
|---|---|---|---|
| Hugging Face | nvidia/Cosmos-Reason2-8B | huggingface.co/nvidia/Cosmos-Reason2-8B | — |
| Hugging Face | nvidia/Cosmos-Reason2-2B | huggingface.co/nvidia/Cosmos-Reason2-2B | — |
Notable capabilities (2)
- Physical-AI reasoning VLM: Spatio-temporal video reasoning, 2D/3D point and box localization, robot planning; 8B beats base Qwen3-VL-8B on robotics (56.90 vs 53.08) and self-driving (67.85 vs 46.38) evals per model card. source
- Backbone for GR00T N1.7 (found after launch): Cosmos-Reason2-2B is the VLM backbone of Isaac GR00T N1.7. source
Other NVIDIA models
NVIDIA Nemotron 3.5 Lightning (30B-A3B) · NVIDIA NemotronLabs VoiceChat 11B (and PersonaPlex-7B) · NVIDIA Nemotron 3 Ultra (550B-A55B) · Cosmos 3 (Nano / Super) · NVIDIA Nemotron 3 Nano Omni (30B-A3B Reasoning) · Isaac GR00T N1.7 · NVIDIA Nemotron 3 Super (120B-A12B) · NVIDIA MagpieTTS Multilingual 357M · NVIDIA Parakeet / Canary / Nemotron Speech ASR (open) · Isaac GR00T N2 · Cosmos Predict 2.5 / Transfer 2.5 · Isaac GR00T N1 / N1.5 / N1.6