Quantization format
UD_Q5_K_S
UD_Q5_K_S has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 28 files measured
Nominal bpw
—
from the block layout
Measured average
6.255
28 files
Range
5.485–14.200
varies by architecture
File sizes
5.59 GiB+
up to 664.21 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 23.23 GiB | 5.550 | — |
| gemma-4-26B-A4B-it | 17.56 GiB | 5.681 | — |
| Qwen-AgentWorld-35B-A3B | 23.23 GiB | 5.757 | — |
| LTX-2.3 | 15.09 GiB | 14.091 | — |
| GLM-5.2 | 491.05 GiB | 5.599 | — |
| Ornith-1.0-35B | 23.23 GiB | 5.757 | — |
| Qwen3.5-122B-A10B | 82.25 GiB | 5.648 | — |
| Qwen3-Coder-Next | 51.99 GiB | 5.605 | — |
| Laguna-S-2.1 | 76.98 GiB | 5.625 | — |
| Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 | 23.10 GiB | 6.011 | — |
| Step-3.7-Flash | 128.59 GiB | 5.486 | — |
| LFM2.5-8B-A1B | 5.59 GiB | 5.674 | — |
| Ornith-1.0-397B | 258.42 GiB | 5.594 | — |
| Qwen3.5-397B-A17B | 257.56 GiB | 5.484 | — |
| KAT-Coder-V2.5-Dev | 23.79 GiB | 5.895 | — |
| North-Mini-Code-1.0 | 20.13 GiB | 5.673 | — |
| MiniMax-M2.7 | 148.12 GiB | 5.563 | — |
| MiniMax-M3 | 278.24 GiB | 5.597 | — |
| NVIDIA-Nemotron-3-Super-120B-A12B-BF16 | 83.56 GiB | 5.806 | — |
| Mistral-Small-4-119B-2603 | 77.82 GiB | 5.598 | — |
| Huihui-gemma-4-26B-A4B-it-abliterated | 17.48 GiB | 5.657 | — |
| MiMo-V2.5 | 201.45 GiB | 5.568 | — |
| GLM-5.1 | 489.83 GiB | 5.581 | — |
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | 364.80 GiB | 5.590 | — |
| MiMo-V2.5-Pro | 664.21 GiB | 5.576 | — |
●From the filewhat these mean