3-bit Open on Hugging Face
Parameters
15B
Weights (est.)
≈ 6.4 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-lm on an Apple silicon Mac:
pip install mlx-lmmlx_lm.generate --model mlx-community/phi-4-3bit --prompt "Hello"
Details
- Base model
- microsoft/phi-4 (quantized) · phi 4 on the radar →
- Vendor
- Microsoft · Phi
- Task
- Text generation
- Quantisation
- 3-bit
- Licence
- mit
- Downloads
- 133 in the last 30 days · 840 all time · 0 likes
- Published
- 2024-12-16 · updated 2025-01-12 (1y ago)
- Languages
- en
- Tags
- phi3, math, code, chat, custom_code
Other MLX builds of microsoft/phi-4
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| phi-4-2bit | 2-bit | ≈ 4.6 GB | 8 GB+ | 72 |
| phi-4-3bit (this one) | 3-bit | ≈ 6.4 GB | 16 GB+ | 133 |
| phi-4-4bit | 4-bit | ≈ 8.2 GB | 16 GB+ | 1.4K |
| phi-4-6bit | 6-bit | ≈ 11.9 GB | 16 GB+ | 182 |
| phi-4-8bit | 8-bit | ≈ 15.6 GB | 24 GB+ | 777 |
| phi-4-bf16 | bf16 | ≈ 29.3 GB | 48 GB+ | 141 |