Quantization format
UD_IQ4_XS
UD_IQ4_XS has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 36 files measured
Nominal bpw
—
from the block layout
Measured average
3.989
36 files
Range
3.594–4.947
varies by architecture
File sizes
1.02 GiB+
up to 461.08 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 16.51 GiB | 3.945 | — |
| gemma-4-26B-A4B-it | 12.66 GiB | 4.098 | — |
| DeepSeek-V4-Flash | 128.43 GiB | 3.792 | — |
| Qwen-AgentWorld-35B-A3B | 16.56 GiB | 4.105 | — |
| GLM-5.2 | 340.22 GiB | 3.879 | — |
| Ornith-1.0-35B | 16.56 GiB | 4.105 | — |
| Kimi-K2.7-Code | 461.08 GiB | 3.741 | — |
| Qwen3.5-122B-A10B | 56.09 GiB | 3.852 | — |
| Qwen3-Coder-Next | 35.79 GiB | 3.859 | — |
| Qwen3.5-35B-A3B | 16.29 GiB | 3.891 | — |
| Laguna-S-2.1 | 53.61 GiB | 3.917 | — |
| Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 | 18.19 GiB | 4.734 | — |
| Step-3.7-Flash | 88.79 GiB | 3.788 | — |
| LFM2.5-8B-A1B | 3.97 GiB | 4.029 | — |
| DeepSeek-R1-Distill-Qwen-1.5B | 1.02 GiB | 4.947 | — |
| Ornith-1.0-397B | 178.91 GiB | 3.873 | — |
| Qwen3.5-397B-A17B | 176.70 GiB | 3.763 | — |
| KAT-Coder-V2.5-Dev | 16.96 GiB | 4.203 | — |
| North-Mini-Code-1.0 | 14.18 GiB | 3.997 | — |
| MiniMax-M2.7 | 100.97 GiB | 3.792 | — |
| MiniMax-M3 | 193.31 GiB | 3.888 | — |
| NVIDIA-Nemotron-3-Super-120B-A12B-BF16 | 60.06 GiB | 4.173 | — |
| Mistral-Small-4-119B-2603 | 54.13 GiB | 3.895 | — |
| Huihui-gemma-4-26B-A4B-it-abliterated | 12.50 GiB | 4.044 | — |
| MiMo-V2.5 | 139.18 GiB | 3.847 | — |
●From the filewhat these mean