6-bit Open on Hugging Face
Parameters
2.1B
Weights (est.)
≈ 2.2 GB
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.
Details
- Base model
- Qwen/Qwen3-VL-2B-Instruct (finetune)
- Vendor
- Qwen
- Task
- Reranking
- Quantisation
- 6-bit
- Licence
- apache-2.0
- Downloads
- 79 in the last 30 days · 333 all time · 0 likes
- Published
- 2026-03-14 · updated 2026-03-14 (8mo ago)
- Tags
- qwen3_vl, image-text-to-text, multimodal rerank, text rerank
Other MLX builds of Qwen/Qwen3-VL-2B-Instruct
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3-VL-Embedding-2B-4bit | 4-bit | ≈ 1.8 GB | 8 GB+ | 338 |
| Qwen3-VL-Reranker-2B-4bit | 4-bit | ≈ 1.8 GB | 8 GB+ | 158 |
| Qwen3-VL-Reranker-2B-mxfp4 | MXFP4 | ≈ 1.8 GB | 8 GB+ | 81 |
| Qwen3-VL-Embedding-2B-mxfp4 | MXFP4 | ≈ 1.8 GB | 8 GB+ | 80 |
| Qwen3-VL-Reranker-2B-5bit | 5-bit | ≈ 2.0 GB | 8 GB+ | 78 |
| Qwen3-VL-Embedding-2B-5bit | 5-bit | ≈ 2.0 GB | 8 GB+ | 33 |
| Qwen3-VL-Reranker-2B-6bit (this one) | 6-bit | ≈ 2.2 GB | 8 GB+ | 79 |
| Qwen3-VL-Embedding-2B-6bit | 6-bit | ≈ 2.2 GB | 8 GB+ | 64 |
| Qwen3-VL-Embedding-2B-8bit | 8-bit | ≈ 2.6 GB | 8 GB+ | 930 |
| Qwen3-VL-Reranker-2B-8bit | 8-bit | ≈ 2.6 GB | 8 GB+ | 149 |
| Qwen3-VL-Reranker-2B-mxfp8 | MXFP8 | ≈ 2.6 GB | 8 GB+ | 80 |
| Qwen3-VL-Embedding-2B-mxfp8 | MXFP8 | ≈ 2.6 GB | 8 GB+ | 71 |
| Qwen3-VL-Embedding-2B-bf16 | bf16 | ≈ 4.3 GB | 8 GB+ | 130 |
| Qwen3-VL-Reranker-2B-bf16 | bf16 | ≈ 4.3 GB | 8 GB+ | 93 |
| Qwen3-VL-Reranker-2B-nvfp4 | — | ≈ 7.7 GB | 16 GB+ | 75 |
| Qwen3-VL-Embedding-2B-nvfp4 | — | ≈ 7.7 GB | 16 GB+ | 30 |