MXFP4 Open on Hugging Face
Parameters
597M
Weights (est.)
≈ 336 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.
Details
- Base model
- Qwen/Qwen3-0.6B (finetune) · Qwen3 0.6B on the radar →
- Vendor
- Qwen
- Task
- Reranking
- Quantisation
- MXFP4
- Licence
- cc-by-nc-4.0
- Downloads
- 142 in the last 30 days · 1.2K all time · 3 likes
- Published
- 2026-02-09 · updated 2026-02-09 (9mo ago)
- Languages
- multilingual
- Tags
- qwen3, feature-extraction, reranker, custom_code
Other MLX builds of Qwen/Qwen3-0.6B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Qwen3-0.6B-3bit | 3-bit | ≈ 261 MB | 8 GB+ | 72 |
| Qwen3-0.6B-4bit | 4-bit | ≈ 335 MB | 8 GB+ | 44K |
| Qwen3-0.6B-4bit-DWQ | 4-bit DWQ | ≈ 335 MB | 8 GB+ | 270 |
| Qwen3-0.6B-4bit-AWQ | 4-bit | ≈ 335 MB | 8 GB+ | 117 |
| Qwen3-0.6B-4bit-DWQ-053125 | 4-bit DWQ | ≈ 335 MB | 8 GB+ | 104 |
| Qwen3-0.6B-4bit-DWQ-05092025 | 4-bit DWQ | ≈ 335 MB | 8 GB+ | 53 |
| jina-reranker-v3-4bit-mxfp4 (this one) | MXFP4 | ≈ 336 MB | 8 GB+ | 142 |
| Qwen3-0.6B-6bit | 6-bit | ≈ 484 MB | 8 GB+ | 80 |
| Qwen3-0.6B-8bit | 8-bit | ≈ 633 MB | 8 GB+ | 40K |
| Qwen3-0.6B-bf16 | bf16 | ≈ 1.2 GB | 8 GB+ | 986 |
| Qwen3-0.6B | — | — | — | 0 |