bf16 Open on Hugging Face
Parameters
425M
Weights (est.)
≈ 849 MB
Mac memory (est.)
8 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/Qwen3.8-27B-MTP-bf16 --prompt "Hello"
Details
- Base model
- Qwen/Qwen3.8-27B (finetune) · Qwen3.8 27B on the radar →
- Vendor
- Qwen
- Task
- Text generation
- Quantisation
- bf16
- Licence
- apache-2.0
- Downloads
- 875 in the last 30 days · 12K all time · 13 likes
- Published
- 2026-08-14 · updated 2026-08-14 (2mo ago)
- Languages
- en
- Tags
- qwen3_5_mtp, mlx-vlm, qwen3.8, qwen3.8-27b, qwen, speculative-decoding, draft-model, bf16
Other MLX builds of Qwen/Qwen3.8-27B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3.8-27B-MTP-4bit | 4-bit | ≈ 239 MB | 8 GB+ | 7.3K |
| Qwen3.8-27B-MTP-nvfp4 | 4-bit | ≈ 239 MB | 8 GB+ | 337 |
| Qwen3.8-27B-MTP-mxfp4 | MXFP4 | ≈ 239 MB | 8 GB+ | 270 |
| Qwen3.8-27B-MTP-8bit | 8-bit | ≈ 451 MB | 8 GB+ | 2.3K |
| Qwen3.8-27B-MTP-mxfp8 | MXFP8 | ≈ 451 MB | 8 GB+ | 365 |
| Qwen3.8-27B-MTP-bf16 (this one) | bf16 | ≈ 849 MB | 8 GB+ | 875 |
| Qwen3.8-27B-4bit | 4-bit | ≈ 16.1 GB | 24 GB+ | 177K |
| Qwen3.8-27B-OptiQ-4bit | 4-bit | ≈ 16.1 GB | 24 GB+ | 8.8K |
| Qwen3.8-27B-mxfp4 | MXFP4 | ≈ 16.1 GB | 24 GB+ | 2.7K |
| Qwen3.8-27B-nvfp4 | 4-bit | ≈ 16.1 GB | 24 GB+ | 1.1K |
| Qwen3.8-27B-GPTQ-4bit | 4-bit | ≈ 16.1 GB | 24 GB+ | 891 |
| Qwen3.8-27B-8bit | 8-bit | ≈ 29.5 GB | 48 GB+ | 27K |
| Qwen3.8-27B-mxfp8 | MXFP8 | ≈ 29.5 GB | 48 GB+ | 591 |
| Qwen3.8-27B-bf16 | bf16 | ≈ 54.7 GB | 64 GB+ | 5.9K |