Google · First MLX build 3 Jun 2026
gemma-4-12B
MLX
TextVisionAudio apache-2.0 Trending
A 12B parameter any to any model from Google, with 8 MLX builds, from 4-bit to bf16.
- Parameters
- 12B From Hugging Face
- Quantisation
- 4-bit gemma-4-12B-4bit
- Memory needed
- 10 GB Estimate, unified memory
- Download
- 6.7 GB Weights, estimated
- Downloads
- 899 This build, last 30 days
- Task
- Any to any Text, Vision, Audio
- MLX builds
- 8 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 10 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · Hugging Face CLI$ pip install -U huggingface_hub
$ hf download mlx-community/gemma-4-12B-4bitMLX builds
8 on mlx-community| gemma-4-12B-4bit | 4-bit | 6.7 GB | 10 GB | 899 |
|---|---|---|---|---|
| gemma-4-12B-mxfp4 | MXFP4 | 6.7 GB | 10 GB | 93 |
| gemma-4-12B-nvfp4 | 4-bit | 6.7 GB | 10 GB | 84 |
| gemma-4-12B-5bit | 5-bit | 8.2 GB | 12 GB | 73 |
| gemma-4-12B-6bit | 6-bit | 9.7 GB | 13 GB | 127 |
| gemma-4-12B-8bit | 8-bit | 12.7 GB | 17 GB | 219 |
| gemma-4-12B-mxfp8 | MXFP8 | 12.7 GB | 17 GB | 85 |
| gemma-4-12B-bf16 | bf16 | 23.9 GB | 29 GB | 144 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.