> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nemo/automodel/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/automodel/_mcp/server.

# Qwen2.5-Omni-3B

> Use Qwen2.5-Omni-3B with NeMo AutoModel for multimodal fine-tuning, with documented checkpoints, runnable recipes, setup guidance, and model reference details.

[Qwen2.5-Omni](https://qwenlm.github.io/blog/qwen2.5-omni/) is Alibaba Cloud's omnimodal model supporting text, image, audio, and video inputs in a single unified architecture with a dense language backbone. NeMo AutoModel onboards the **Thinker** stack for audio understanding tasks such as automatic speech recognition (ASR).

Set up NeMo AutoModel with the [latest container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/nemo-automodel) or follow the [installation instructions](/get-started/installation).

## Fine-Tune Qwen2.5-Omni-3B

From the repository root, run:

```bash
uv run automodel --nproc-per-node=8 examples/audio_finetune/qwen2_5_omni_asr/ami_sft_3b.yaml
```

## Choose a Workflow

| Goal                                                | Start Here                                                                                                                                          |
| --------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| ASR SFT - Qwen2.5-Omni 3B on the AMI meeting corpus | Use [ami\_sft\_3b.yaml](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/audio_finetune/qwen2_5_omni_asr/ami_sft_3b.yaml). Dataset: AMI. |

## Model Reference

### Model Architecture

| Property                  | Value                                 |
| ------------------------- | ------------------------------------- |
| Task                      | Omnimodal (Text/Image/Audio/Video)    |
| Architecture              | `Qwen2_5OmniForConditionalGeneration` |
| Parameters                | 3B                                    |
| Hugging Face Organization | [Qwen](https://huggingface.co/Qwen)   |

The registry wires the Qwen2.5-Omni Thinker backbone under the following architecture keys:

* `Qwen2_5OmniForConditionalGeneration`
* `Qwen2_5OmniModel`
* `Qwen2_5OmniThinkerForConditionalGeneration`

### Available Models

| Model           | HF ID                                                                 |
| --------------- | --------------------------------------------------------------------- |
| Qwen2.5-Omni 3B | [`Qwen/Qwen2.5-Omni-3B`](https://huggingface.co/Qwen/Qwen2.5-Omni-3B) |

## Related Resources

* [VLM Fine-Tuning Guide](/recipes-e2e-examples/gemma-3-3n)