---
license: apache-2.0
library_name: llama.cpp
pipeline_tag: image-text-to-text
tags:
  - gguf
  - qwen3-vl
  - vision-language
  - ollama
  - llama.cpp
  - jetson
---

# Open3DVQA Qwen3-VL 4B Distilled Q4_K_M GGUF

This repository contains the deployment files for the distilled Open3DVQA
Qwen3-VL 4B student model.

## Files

```text
student_4b_merged-Q4_K_M.gguf       Q4_K_M language model, about 2.4 GB
mmproj-student_4b_merged-f16.gguf   F16 vision encoder/projector, about 798 MB
Modelfile.ollama                    Ollama import configuration
CODEX_ORIN_DEPLOY_GUIDE.md          Detailed Jetson Orin NX instructions
```

Both GGUF files are required for image inference.

## Ollama

```bash
ollama create open3dvqa-qwen3vl:4b-q4km -f Modelfile.ollama
```

Use Ollama's `/api/chat` endpoint with base64 image data in
`messages[].images`.

See `CODEX_ORIN_DEPLOY_GUIDE.md` for JetPack 5 deployment, checksums, GPU
verification, API examples, and troubleshooting.