Quantization format
IQ3_XS
IQ3_XS has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 788 files measured
Nominal bpw
—
from the block layout
Measured average
3.562
788 files
Range
1.297–13.839
varies by architecture
File sizes
0.03 GiB+
up to 434.58 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 15.94 GiB | 3.807 | — |
| gemma-4-26B-A4B-it | 11.58 GiB | 3.748 | — |
| gemma-4-12B-it | 5.15 GiB | 3.696 | — |
| Qwen3.5-4B | 2.24 GiB | 4.127 | — |
| Hy3 | 128.00 GiB | 3.680 | — |
| gemma-4-31B-it | 12.89 GiB | 3.541 | — |
| gemma-4-E2B-it | 2.89 GiB | 4.842 | — |
| Qwen3-8B | 3.38 GiB | 3.542 | — |
| Qwen3.5-0.8B | 0.46 GiB | 4.560 | — |
| Llama-3.1-8B-Instruct | 3.28 GiB | 3.506 | — |
| Ornith-1.0-35B | 15.10 GiB | 3.743 | — |
| ThinkingCap-Qwen3.6-27B | 12.41 GiB | 3.898 | — |
| Qwythos-9B-v2 | 4.37 GiB | 3.893 | — |
| Qwen3.5-122B-A10B | 55.13 GiB | 3.786 | — |
| Qwen3-Coder-Next | 30.76 GiB | 3.317 | — |
| Qwen2.5-32B-Instruct | 12.76 GiB | 3.346 | — |
| Qwen3.5-35B-A3B | 15.94 GiB | 3.807 | — |
| Qwen2.5-Coder-7B-Instruct | 3.12 GiB | 3.515 | — |
| Qwen2.5-7B-Instruct | 3.12 GiB | 3.515 | — |
| Qwen3-0.6B | 0.35 GiB | 4.040 | — |
| Qwen3-VL-2B-Instruct | 0.78 GiB | 3.137 | — |
| Jan-v3-4B-base-instruct | 1.85 GiB | 3.593 | — |
| Qwen2.5-1.5B-Instruct | 0.68 GiB | 3.792 | — |
| Qwen2.5-VL-7B-Instruct | 3.12 GiB | 3.228 | — |
| Ornith-1.0-9B | 4.25 GiB | 3.967 | — |
●From the filewhat these mean