# Cosmos 3 Nano Reasoner

> A vision language model that excels in understanding the physical world using structured reasoning on videos or images. Built on NVIDIA's Cosmos foundation for physical AI, it provides dense, step-by-step reasoning about visual scenes and temporal sequences. Served via NVIDIA NIM.

**ID**: `nvidia/cosmos3-nano-reasoner`  
**Creator**: nvidia  
**Category**: multimodal  
**Context**: 128K tokens  
**Status**: Disabled — no longer routable or listed; kept for history/usage reference.  
**Reason**: NVIDIA has not exposed this on the hosted API yet (integrate.api.nvidia.com 400s); it ships as self-host NIM + HF weights only.  
**Input**: $0  
**Output**: $0  
**Released**: 2026-06-04  
**Web page**: https://anyrouter.dev/model/nvidia/cosmos3-nano-reasoner

**Input modalities**: text, image, video  
**Output modalities**: text  

**Tokenizer**: Nemotron

**Capabilities**: chat, streaming, vision, reasoning

**Supported parameters**: max_tokens, temperature, top_p, stop, tools, tool_choice, response_format

## Usage

**Endpoint**: `POST https://anyrouter.dev/api/v1/chat/completions`

```bash
curl https://anyrouter.dev/api/v1/chat/completions \
  -H "Authorization: Bearer $ANYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "nvidia/cosmos3-nano-reasoner",
  "messages": [
    {
      "role": "user",
      "content": "Say hi in 3 words."
    }
  ]
}'
```

## Providers

| Provider | Input | Output |
| --- | --- | --- |
| nvidia | $0 | $0 |
| nvidia-byok | $0 | $0 |
| openrouter-byok | $0 | $0 |

## Source

- Upstream docs: https://build.nvidia.com/nvidia/cosmos3-nano-reasoner