Quantization format
UD_IQ1_S
UD_IQ1_S has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 157 files measured
Nominal bpw
—
from the block layout
Measured average
2.160
157 files
Range
1.387–4.457
varies by architecture
File sizes
0.20 GiB+
up to 265.74 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| DeepSeek-V4-Flash | 76.87 GiB | 2.270 | — |
| Qwen3-Coder-30B-A3B-Instruct | 8.30 GiB | 2.336 | — |
| Qwen3-VL-30B-A3B-Instruct | 8.46 GiB | 2.340 | — |
| Llama-3.2-1B-Instruct | 0.39 GiB | 2.729 | — |
| Qwen3-8B | 2.12 GiB | 2.222 | — |
| Qwen3-4B | 1.01 GiB | 2.154 | — |
| GLM-5.2 | 201.15 GiB | 2.294 | — |
| Llama-3.1-8B-Instruct | 2.02 GiB | 2.156 | — |
| Ornith-1.0-35B | 9.80 GiB | 2.429 | — |
| Llama-3.2-3B-Instruct | 0.85 GiB | 2.272 | — |
| Qwen3-Coder-Next | 20.03 GiB | 2.160 | — |
| gemma-3-1b-it | 0.52 GiB | 4.457 | — |
| Qwen3-14B | 3.56 GiB | 2.073 | — |
| Qwen3-VL-4B-Instruct | 1.01 GiB | 1.953 | — |
| Qwen3-0.6B | 0.20 GiB | 2.285 | — |
| Qwen3-VL-2B-Instruct | 0.50 GiB | 2.022 | — |
| Qwen2.5-VL-7B-Instruct | 1.93 GiB | 2.001 | — |
| Qwen3-30B-A3B | 8.42 GiB | 2.369 | — |
| Qwen3-1.7B | 0.50 GiB | 2.118 | — |
| Laguna-S-2.1 | 31.45 GiB | 2.298 | — |
| Qwen3-4B-Instruct-2507 | 1.01 GiB | 2.154 | — |
| gemma-3-12b-it | 2.85 GiB | 2.007 | — |
| GLM-4.7-Flash | 8.61 GiB | 2.370 | — |
| Qwen3-30B-A3B-Instruct-2507 | 8.42 GiB | 2.370 | — |
| DeepSeek-R1-0528-Qwen3-8B | 2.11 GiB | 2.216 | — |
●From the filewhat these mean