Google · First MLX build 15 Apr 2025 · In Core AI since 8 Jun 2026
gemma-3-4b-it
MLXCore AI
TextVision gemma
A 4.3B parameter vision and text model from Google, with 6 MLX builds, from 4-bit DWQ to bf16 and Apple Core AI export recipes for macOS.
- Parameters
- 4.3B From Hugging Face
- Quantisation
- 4-bit DWQ gemma-3-4b-it-4bit-DWQ
- Memory needed
- 5 GB Estimate, unified memory
- Download
- 2.6 GB Weights, estimated
- Downloads
- 98 This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 6 On mlx-community
- Devices
- Mac Apple silicon
- Parameters
- 4.3B From Hugging Face
- Compression
- 4-bit Apple’s preset
- Memory needed
- 5 GB Estimate from parameters
- Context
- 128K tokens Fixed at export
- Precision
- BF16 Compute precision
- Platforms
- macOS Mac
- Export presets
- 1 In Apple’s registry
- Listed by Apple
- 8 Jun 2026 First registry commit
Memory
Will it fit?Needs about 5 GB at 4-bit DWQ
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 · mlx-lm$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/gemma-3-4b-it-4bit-DWQ --prompt "Hello"Memory
Will it fit?Needs about 5 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 parameter count at Apple’s compression setting. Apple does not publish package sizes.
Export it
Terminal · uv$ git clone https://github.com/apple/coreai-models
$ cd coreai-models
$ uv run coreai.llm.export gemma3-4b-it
Apple publishes the recipe, not a ready-made package: these commands convert the original weights to a Core AI .aimodel on your Mac. See the repository for requirements.
MLX builds
6 on mlx-community| gemma-3-4b-it-4bit-DWQ | 4-bit DWQ | 2.6 GB | 5 GB | 98 |
|---|---|---|---|---|
| gemma-3-4b-it-5bit | 5-bit | 2.7 GB | 5 GB | 46 |
| gemma-3-4b-it-qat-3bit | 3-bit | 2.8 GB | 6 GB | 59 |
| gemma-3-4b-it-qat-6bit | 6-bit | 4.5 GB | 8 GB | 29 |
| gemma-3-4b-it-qat-8bit | 8-bit | 5.7 GB | 9 GB | 33 |
| gemma-3-4b-it-qat-bf16 | bf16 | 9.9 GB | 14 GB | 32 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.
Core AI export presets
From Apple’s registry| Preset | Platform | Compression | Precision | Context |
|---|---|---|---|---|
| gemma3-4b-it | macOS | 4-bit | BF16 | 128K |