bf16 Open on Hugging Face
Parameters
31B
Weights (est.)
≈ 61.1 GB
Mac memory (est.)
96 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/Qwen3-30B-A3B-bf16 --prompt "Hello"
Details
- Base model
- Qwen/Qwen3-30B-A3B (finetune)
- Vendor
- Qwen
- Task
- Text generation
- Quantisation
- bf16
- Licence
- apache-2.0
- Downloads
- 91 in the last 30 days · 6.5K all time · 2 likes
- Published
- 2025-04-28 · updated 2025-04-28 (1y ago)
- Tags
- qwen3_moe
Other MLX builds of Qwen/Qwen3-30B-A3B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3-30B-A3B-3bit | 3-bit | ≈ 13.4 GB | 24 GB+ | 185 |
| Qwen3-30B-A3B-4bit | 4-bit | ≈ 17.2 GB | 24 GB+ | 7.9K |
| Qwen3-30B-A3B-4bit-DWQ | 4-bit DWQ | ≈ 17.2 GB | 24 GB+ | 559 |
| Qwen3-30B-A3B-4bit-DWQ-0508 | 4-bit DWQ | ≈ 17.2 GB | 24 GB+ | 136 |
| Qwen3-30B-A3B-mixed-3-4bit | 4-bit | ≈ 17.2 GB | 24 GB+ | 72 |
| Qwen3-30B-A3B-Instruct-2507-4bit-ODWQ | 4-bit | ≈ 17.2 GB | 24 GB+ | 70 |
| Qwen3-30B-A3B-4bit-DWQ-10072025 | 4-bit DWQ | ≈ 17.2 GB | 24 GB+ | 56 |
| Qwen3-30B-A3B-4bit-DWQ-05082025 | 4-bit DWQ | ≈ 17.2 GB | 24 GB+ | 53 |
| Qwen3-30B-A3B-4bit-DWQ-053125 | 4-bit DWQ | ≈ 17.2 GB | 24 GB+ | 47 |
| Qwen3-30B-A3B-6bit | 6-bit | ≈ 24.8 GB | 32 GB+ | 74 |
| Qwen3-30B-A3B-8bit | 8-bit | ≈ 32.4 GB | 48 GB+ | 933 |
| Qwen3-30B-A3B-bf16 (this one) | bf16 | ≈ 61.1 GB | 96 GB+ | 91 |