OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| MMMU | 68 | College-level multimodal understanding + reasoning | Source | September 21, 2026 |
| AI2D | 87.11 | Question answering over science diagrams | Source | September 21, 2026 |
| ChartQA | 89.72 | Question answering over charts and plots | Source | September 21, 2026 |
| DocVQA | 94.39 | Question answering over document images | Source | September 21, 2026 |
| OCRBench | 85.6 | Text recognition + reasoning over images of text | Source | September 21, 2026 |
| Intelligence | 7.5 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 22, 2026 |
| Math | 75 | Composite score across standardised mathematics evaluations, measured independently by Artificial Analysis | Source | September 22, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Nemotron Nano 12B V2 VL This page | ||||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
Yes, Nemotron Nano 12B V2 VL can stream responses as they are generated.
Yes, Nemotron Nano 12B V2 VL supports both function calling and structured outputs.
Nemotron Nano 12B V2 VL was built by NVIDIA.
The model is accessed via the v1/chat/completions endpoint.
