MLX community ships 4-bit Gemma 4 E2B VLM for Apple silicon
TL;DR
- A 4-bit MLX conversion of Google's Gemma 4 E2B-it vision model fits into a 3.55 GB file targeting Apple silicon.
- The listing reports 38,478 downloads in the last month and ships under Google's Gemma license, not Apache 2.0.
- The card exposes the model as image-text-to-text, invoked through mlx_vlm.generate with a prompt and image path.
The MLX community has published a 4-bit conversion of Google's Gemma 4 E2B-it vision-language model, packing the 5-billion-parameter instruction-tuned model into a 3.55 GB file targeted at Apple silicon. The Hugging Face listing reports 38,478 downloads in the last month and names LM Studio, Pi, Hermes Agent, Atomic Chat and OpenClaw as local-app integration targets.
The port was built with `mlx_vlm.convert` off `google/gemma-4-E2B-it` and inherits Google's Gemma license. The card describes itself as an "MLX conversion of google/gemma-4-E2B-it for Apple silicon with 4-bit quantization" and exposes the model as image-text-to-text, invoked via `mlx_vlm.generate` with a prompt and an image path.
No benchmark numbers appear on the page. Two Spaces on Hugging Face and four discussion threads are the only community signal so far.
Shared on Bluesky by 1 AI expert
Originally reported by huggingface.co
Read the original article →Original headline: mlx-community/gemma-4-e2b-it-4bit · Hugging Face