SuperagenticAI/gemma-4-31b-it-4bit-mlx

Gemma 4 31B IT 4bit MLX

This repository contains an MLX 4-bit quantized conversion of google/gemma-4-31B-it, published by Superagentic AI.

Base model: https://huggingface.co/google/gemma-4-31B-it

While model converted, MLX LM wasnt got the Gemma4 Support for the text model uploaded using the local patch of the mlx-lm .. Someone create PR and Support for Gemma for still not merged yet

Notes

• Converted to MLX format for Apple Silicon inference.
• Quantized to approximately 4.5 bits per weight.
• This is a conversion release, not a Superagentic AI fine-tune.
• All original Gemma license terms and restrictions continue to apply.

Usage

Load with mlx-lm on Apple Silicon:

mlx_lm.generate \
  --model SuperagenticAI/gemma-4-31b-it-4bit-mlx \
  --prompt "Explain what is Agentic AI."