8-bit Open on Hugging Face
Parameters
5.0B
Weights (est.)
≈ 5.7 GB
Mac memory (est.)
16 GB+
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/gemma-3-4b-it-qat-8bit --image photo.jpg --prompt "Describe this image"
Details
- Base model
- google/gemma-3-4b-it (finetune)
- Vendor
- Google · Gemma
- Task
- Vision + text
- Quantisation
- 8-bit
- Licence
- gemma
- Downloads
- 33 in the last 30 days · 919 all time · 0 likes
- Published
- 2025-04-15 · updated 2025-04-18 (1y ago)
- Tags
- gemma3, gemma, google
Other MLX builds of google/gemma-3-4b-it
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| gemma-3-4b-it-4bit-DWQ | 4-bit DWQ | ≈ 2.6 GB | 8 GB+ | 100 |
| gemma-3-4b-it-5bit | 5-bit | ≈ 2.7 GB | 8 GB+ | 46 |
| gemma-3-4b-it-qat-3bit | 3-bit | ≈ 2.8 GB | 8 GB+ | 57 |
| Gemma-SEA-LION-v4-4B-VL-mlx-3bit | 3-bit | ≈ 2.8 GB | 8 GB+ | 22 |
| gemma-3-4b-it-qat-6bit | 6-bit | ≈ 4.5 GB | 8 GB+ | 29 |
| gemma-3-4b-it-qat-8bit (this one) | 8-bit | ≈ 5.7 GB | 16 GB+ | 33 |
| gemma-3-4b-it-qat-bf16 | bf16 | ≈ 9.9 GB | 16 GB+ | 32 |