bf16 Open on Hugging Face
Parameters
73B
Weights (est.)
≈ 147 GB
Mac memory (est.)
192 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-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/QVQ-72B-Preview-bf16 --image photo.jpg --prompt "Describe this image"
Details
- Base model
- Qwen/Qwen2-VL-72B (finetune)
- Vendor
- Qwen
- Task
- Vision + text
- Quantisation
- bf16
- Licence
- qwen
- Downloads
- 20 in the last 30 days · 338 all time · 3 likes
- Published
- 2024-12-24 · updated 2024-12-24 (1y ago)
- Languages
- en
- Tags
- qwen2_vl, chat
Other MLX builds of Qwen/Qwen2-VL-72B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| QVQ-72B-Preview-3bit | 3-bit | ≈ 32.1 GB | 48 GB+ | 22 |
| QVQ-72B-Preview-4bit | 4-bit | ≈ 41.3 GB | 64 GB+ | 34 |
| QVQ-72B-Preview-6bit | 6-bit | ≈ 59.6 GB | 96 GB+ | 20 |
| QVQ-72B-Preview-8bit | 8-bit | ≈ 78.0 GB | 96 GB+ | 25 |
| QVQ-72B-Preview-bf16 (this one) | bf16 | ≈ 147 GB | 192 GB+ | 20 |