Cosmos Reason2 8B
Vision language model for physical-world understanding with structured reasoning on video and images
Model facts
- Context window
- 131K tokens
- Maximum output
- 16K tokens
- Cheapest paid input
- free per 1M tokens
- Cheapest paid output
- free per 1M tokens
- Open weights
- yes
- Providers
- 1
Capabilities and modalities
reasoning, tool calling, attachments, temperature control, open weights, text, image, video.
Local hardware estimate
At 4K context: Q4_K_M 6.3 GiB, Q8_0 10.4 GiB, F16 18.1 GiB working memory.
- Parameters
- 8.8 billion
- Architecture
- qwen3vl
- Model license
- other
Planning estimate adapted from the MIT-licensed whichllm estimator using metadata from Hugging Face; not a vendor minimum requirement.
Observed price history
- 2026-08-24: free input / free output per 1M tokens
API providers
- Nvidia — model id nvidia/cosmos-reason2-8b; free input / free output per 1M tokens; provider documentation