ABEJA-Qwen2.5-32b-Japanese-v0.1-8bit
mlx-community/ABEJA-Qwen2.5-32b-Japanese-v0.1-8bit
Text generation · transformers
8-bit Open on Hugging Face
Parameters
33B
Weights (est.)
≈ 34.8 GB
Mac memory (est.)
48 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/ABEJA-Qwen2.5-32b-Japanese-v0.1-8bit --prompt "Hello"
Details
- Base model
- abeja/ABEJA-Qwen2.5-32b-Japanese-v0.1 (quantized)
- Vendor
- abeja
- Task
- Text generation
- Quantisation
- 8-bit
- Licence
- apache-2.0
- Downloads
- 106 in the last 30 days · 399 all time · 0 likes
- Published
- 2025-01-27 · updated 2025-01-27 (1y ago)
- Languages
- ja
- Tags
- qwen2
Other MLX builds of abeja/ABEJA-Qwen2.5-32b-Japanese-v0.1
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| ABEJA-Qwen2.5-32b-Japanese-v0.1-4bit | 4-bit | ≈ 18.4 GB | 24 GB+ | 106 |
| ABEJA-Qwen2.5-32b-Japanese-v0.1-8bit (this one) | 8-bit | ≈ 34.8 GB | 48 GB+ | 106 |
| ABEJA-Qwen2.5-32b-Japanese-v0.1 | — | ≈ 65.5 GB | 96 GB+ | 116 |