---
language: en
pipeline_tag: image-text-to-text
library_name: mlx
tags:
- mlx
base_model:
- ibm-granite/granite-vision-4.1-4b
base_model_relation: quantized
---
# Granite-Vision-4.1-4B
See Granite-Vision-4.1 in action: [demonstration videos](https://youtube.com/xcreate)
   
#### Tested with an M3 Ultra 512 GiB using [Inferencer app v1.11.7](https://inferencer.com)
- Vision inference: ~72.5 tokens/s @ 1000 tokens ~6.5 GiB (debug build)

Q9 typically achieves near lossless accuracy in our coding test.

<img src="https://cdn-uploads.huggingface.co/production/uploads/688479d616f1ec82fa645019/1iYIYwBc-DbMEAKy1HOLs.jpeg" style="display:block;margin-top:0;">

##### Quantized with a modified version of [MLX](https://github.com/ml-explore/mlx)
##### For more details see our [demonstration videos](https://youtube.com/xcreate) or visit [granite-vision-4.1-4b](ibm-granite/granite-vision-4.1-4b).

## Disclaimer

We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.