4-bit Open on Hugging Face
Parameters
31B
Weights (est.)
≈ 18.4 GB
Mac memory (est.)
24 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-lm on an Apple silicon Mac:
pip install mlx-lmmlx_lm.generate --model mlx-community/gemma-4-31B-it-OptiQ-4bit --prompt "Hello"
Details
- Base model
- google/gemma-4-31B-it (quantized)
- Vendor
- Google · Gemma
- Task
- Text generation
- Quantisation
- 4-bit
- Licence
- gemma
- Downloads
- 757 in the last 30 days · 30K all time · 15 likes
- Published
- 2026-04-26 · updated 2026-10-18 (24d ago)
- Tags
- gemma4, quantized, mixed-precision, 4bit, 8bit, optiq, apple-silicon, gemma-4
Other MLX builds of google/gemma-4-31B-it
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking-4.6bit-msq | 6-bit | ≈ 6.5 GB | 16 GB+ | 101 |
| gemma-4-31b-it-4bit | 4-bit | ≈ 18.4 GB | 24 GB+ | 8.2K |
| gemma-4-31b-it-mxfp4 | MXFP4 | ≈ 18.4 GB | 24 GB+ | 283 |
| gemma-4-31b-it-nvfp4 | 4-bit | ≈ 18.4 GB | 24 GB+ | 191 |
| gemma-4-31B-it-OptiQ-4bit (this one) | 4-bit | ≈ 18.4 GB | 24 GB+ | 757 |
| gemma-4-31b-it-5bit | 5-bit | ≈ 22.2 GB | 32 GB+ | 169 |
| gemma-4-31b-it-6bit | 6-bit | ≈ 26.1 GB | 32 GB+ | 683 |
| gemma-4-31b-it-8bit | 8-bit | ≈ 33.8 GB | 48 GB+ | 3.4K |
| gemma-4-31b-it-mxfp8 | MXFP8 | ≈ 33.8 GB | 48 GB+ | 178 |
| gemma-4-31b-it-bf16 | bf16 | ≈ 62.5 GB | 96 GB+ | 1.5K |