---
library_name: mlx
license: apache-2.0
license_link: https://ai.google.dev/gemma/docs/gemma_4_license
pipeline_tag: image-text-to-text
base_model: google/gemma-4-26B-A4B
tags:
- mlx
---

# mavis-ai/Gemma4-26B-MoE-Q4

This model is a mixed-policy 4-bit MLX quantized version of the official Google Gemma 4 26B model. No structural modifications or fine-tuning have been applied to the base weights; the package applies MLX quantization while preserving the higher-precision tensor groups indicated by Google's official QAT Q4 GGUF layout.

**💡 Notice for Everyone**
While this repository is hosted primarily as a dedicated engine source for our application ecosystem (R.E.V.I.S.), **you are completely free to download and use this quantized model normally** for your own local MLX projects or workflows. For more details on the base model, please refer directly to the **[Official google/gemma-4-26B-A4B page](https://huggingface.co/google/gemma-4-26B-A4B)**.

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

We quantized and thoroughly tested this model to serve as the local reasoning engine for our project: **R.E.V.I.S.**
*(Note: This specific Q4 model is exactly what we used in our official launch demo!)*

**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 absolute limits of local AI and open-weight models like this one, check out our project!

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

## Quantization

This package follows the quantization boundaries used by Google's official Gemma 4 QAT Q4 GGUF release, adapted for a regular Q4 build:

- Default language tensor quantization: MLX affine Q4
- Token embeddings: MLX affine Q6
- Shared MLP and expert tensors: MLX affine Q4
- Router projection and router scale tensors: BF16
- Vision and multimodal projection components: BF16
- Group size: `64`
- Mode: `affine`

## License

This model is a quantized derivative of Google's Gemma 4 (base model: [google/gemma-4-26B-A4B](https://huggingface.co/google/gemma-4-26B-A4B)), which is released 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:** The base weights have been converted to MLX and quantized with a mixed Q4/Q6/BF16 policy. No fine-tuning or structural modifications were made to the model architecture.