4-bit Open on Hugging Face
Parameters
35B
Weights (est.)
≈ 20.4 GB
Mac memory (est.)
32 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/Agents-A1-OptiQ-4bit --image photo.jpg --prompt "Describe this image"
Details
- Base model
- InternScience/Agents-A1 (quantized)
- Vendor
- InternScience
- Task
- Vision + text
- Quantisation
- 4-bit
- Licence
- apache-2.0
- Downloads
- 111 in the last 30 days · 1.7K all time · 3 likes
- Published
- 2026-07-14 · updated 2026-09-15 (1mo ago)
- Tags
- qwen3_5_moe, quantized, mixed-precision, 4bit, 8bit, optiq, apple-silicon, vision-language, vision, agentic
Other MLX builds of InternScience/Agents-A1
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Agents-A1-3bit | 3-bit | ≈ 16.1 GB | 24 GB+ | 49 |
| Agents-A1-4bit | 4-bit | ≈ 20.4 GB | 32 GB+ | 527 |
| Agents-A1-OptiQ-4bit (this one) | 4-bit | ≈ 20.4 GB | 32 GB+ | 111 |
| Agents-A1-5bit | 5-bit | ≈ 24.7 GB | 32 GB+ | 53 |
| Agents-A1-6bit | 6-bit | ≈ 29.1 GB | 48 GB+ | 90 |
| Agents-A1-8bit | 8-bit | ≈ 37.7 GB | 48 GB+ | 113 |
| Agents-A1-bf16 | bf16 | ≈ 70.2 GB | 96 GB+ | 61 |