mavis-ai/Gemma4-E2B-MLX

This repository contains an MLX-ready full-precision distribution of Google's official Gemma 4 E2B instruction-tuned model, prepared for local inference on Apple Silicon Macs.

Important: this is the same upstream model as google/gemma-4-E2B-it. It is not a new model, not a fine-tune, and not a quantized variant. The only packaging change is conversion to the MLX file layout for local Apple Silicon inference.

Important Notice

This repository is hosted primarily as a dedicated engine source for the R.E.V.I.S. application ecosystem. You are free to download and use this model package for your own local MLX projects or workflows, subject to the Apache License 2.0 and Google's Gemma terms.

For the original model card, architecture details, intended usage, limitations, and evaluation information, refer to the official upstream model:

• Base model: https://huggingface.co/google/gemma-4-E2B-it

Model Identity

This package is intended to match the official Google Gemma 4 E2B instruction-tuned checkpoint in model behavior and weights. MAVIS has not applied additional training, alignment, pruning, merging, adapter injection, architecture changes, or quantization.

The files are redistributed in MLX format so that local MLX / mlx-vlm runtimes can load the model directly on Apple Silicon Macs.

Precision

This is the full MLX build, not a Q4/Q5/Q6/Q8 build:

• Source checkpoint: google/gemma-4-E2B-it
• Weight precision: BF16 where represented by the upstream conversion path
• Quantization: none
• Fine-tuning: none
• Architecture changes: none

If you need a smaller download or lower memory footprint, use the separate MAVIS Q4/Q5/Q6/Q8 repositories instead.

Optimized for R.E.V.I.S. (Local Cognitive OS)

We host and test this model package to serve as a compact local reasoning and judgment engine for R.E.V.I.S.

R.E.V.I.S. is a 100% local Cognitive OS for Multi-Agentic AI. It transforms your Mac devices into a distributed Agentic Swarm via zero-config Wi-Fi clustering, allowing you to run heavy AI workloads like recursive web research, dynamic RAG generation, and multi-step logic without killing single-machine performance.

If you are interested in pushing the limits of local AI and open-weight models, check out our project.

• Official Website: https://mavis-ai.co.jp/revis/
• Watch the 13-min Raw Demo (Multi-node Dynamic RAG): https://x.gd/LxaBF
• Follow our updates on X: https://x.com/mavis_ai_jp

Usage

Install or update the MLX runtime you use for Gemma 4 / multimodal models:

pip install -U mlx mlx-lm mlx-vlm huggingface_hub hf_xet

Download the model:

hf download mavis-ai/Gemma4-E2B-MLX \
  --local-dir ~/Models/mlx/Gemma4-E2B-MLX

Run a quick text generation test:

python -m mlx_vlm.generate \
  --model ~/Models/mlx/Gemma4-E2B-MLX \
  --max-tokens 256 \
  --temperature 0.0 \
  --prompt "Say OK."

Usage Notes

Gemma 4 E2B is the smallest model in this branch and is suited for fast local RAG support, routing, lightweight judgment, and repeated agentic tasks.

For lower memory use, the Q4/Q5/Q6/Q8 MLX variants are available as separate repositories. Those variants intentionally use mixed quantization policies and are not full-precision builds.

License

This repository redistributes an MLX-format derivative of Google's Gemma 4 E2B instruction-tuned release, which is distributed by Google under the Apache License 2.0.

This derivative is likewise distributed under the Apache License 2.0. A copy of the license is included in the LICENSE file in this repository, and can also be found at https://www.apache.org/licenses/LICENSE-2.0.

Modification Notice

Compared with the official Google source checkpoint, this repository applies only the following packaging modification:

The source checkpoint was converted to MLX format for local MLX inference.

No quantization, fine-tuning, additional training, or architecture-level modification has been applied.