metadata
base_model: Qwen/Qwen3-VL-4B-Instruct
base_model_relation: quantized
quantized_by: ArtusDev
license: apache-2.0
pipeline_tag: image-text-to-text
tags:
- exl3
ArtusDev/Qwen_Qwen3-VL-4B-Instruct-EXL3
EXL3 quants of Qwen/Qwen3-VL-4B-Instruct using exllamav3 for quantization.
⚠️ Requires ExLlamaV3 DEV or v0.0.13. ⚠️
Quants
| Quant | BPW | Head Bits | Size (GB) |
|---|---|---|---|
| 2.5_H6 | 2.5 | 6 | 3.05 |
| 3.0_H6 | 3.0 | 6 | 3.28 |
| 3.5_H6 | 3.5 | 6 | 3.51 |
| 4.0_H6 | 4.0 | 6 | 3.73 |
| 4.5_H6 | 4.5 | 6 | 3.96 |
| 5.0_H6 | 5.0 | 6 | 4.19 |
| 6.0_H6 | 6.0 | 6 | 4.64 |
| 8.0_H8 | 8.0 | 8 | 5.65 |
How to Download and Use Quants
You can download quants by targeting specific size using the Hugging Face CLI.
Click for download commands
1. Install huggingface-cli:
pip install -U "huggingface_hub[cli]"huggingface-cli download ArtusDev/Qwen_Qwen3-VL-4B-Instruct-EXL3 --revision "5.0bpw_H6" --local-dir ./EXL3 quants can be run with any inference client that supports EXL3, such as TabbyAPI. Refer to documentation for set up instructions.
Acknowledgements
Made possible with cloud compute from lium.io