---
license: gemma
language:
  - en
base_model: google/gemma-4-26b-a4b-it
tags:
  - tater
  - tater-pick
  - tater-nothink
  - nothink
  - mlx
  - mlx-node
  - quantized
  - awq
  - gemma4
  - moe
  - sliding-window-attention
  - vision-language
  - apple-silicon
  - unsloth-dynamic
  - text-generation
  - conversational
  - en
  - 4-bit
library_name: mlx-node
quantized_by: mlx-node
pipeline_tag: text-generation
model_type: gemma4
---

# Gemma-4-26B-A4B-IT UD-Q4_K_XL MLX - Tater NoThink

This is a Tater NoThink repack of [Brooooooklyn/Gemma-4-26B-A4B-IT-UD-Q4_K_XL-mlx](https://huggingface.co/Brooooooklyn/Gemma-4-26B-A4B-IT-UD-Q4_K_XL-mlx).

## What changed

- The model weights are unchanged.
- `chat_template.jinja` now forces `enable_thinking = false`.
- The model card is tagged for Tater Picks auto-discovery.

## Why

Tater works best with models that answer directly and do not emit hidden reasoning or thinking blocks. This repo keeps the same MLX quantized model but makes the default chat template run in NoThink mode for runtimes that use the embedded template.

## Source

- Source model: [Brooooooklyn/Gemma-4-26B-A4B-IT-UD-Q4_K_XL-mlx](https://huggingface.co/Brooooooklyn/Gemma-4-26B-A4B-IT-UD-Q4_K_XL-mlx)
- Base model: [google/gemma-4-26b-a4b-it](https://huggingface.co/google/gemma-4-26b-a4b-it)
- Format: MLX SafeTensors, UD-Q4_K_XL

## License

This model inherits the [Gemma Terms of Use](https://ai.google.dev/gemma/terms) from the base model.