Quantization format
UD_Q6_K
UD_Q6_K 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
6.740
28 files
Range
6.366–8.138
varies by architecture
File sizes
3.09 GiB+
up to 788.42 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 27.30 GiB | 6.522 | — |
| gemma-4-26B-A4B-it | 21.58 GiB | 6.984 | — |
| Qwen-AgentWorld-35B-A3B | 27.30 GiB | 6.765 | — |
| GLM-5.2 | 582.88 GiB | 6.646 | — |
| Ornith-1.0-35B | 27.30 GiB | 6.765 | — |
| Qwen3.5-122B-A10B | 96.97 GiB | 6.659 | — |
| Qwen3-Coder-Next | 61.27 GiB | 6.606 | — |
| Laguna-S-2.1 | 91.19 GiB | 6.663 | — |
| Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 | 31.28 GiB | 8.138 | — |
| Step-3.7-Flash | 152.12 GiB | 6.489 | — |
| LFM2.5-8B-A1B | 6.60 GiB | 6.697 | — |
| Ornith-1.0-397B | 306.89 GiB | 6.644 | — |
| Qwen3.5-397B-A17B | 304.17 GiB | 6.477 | — |
| KAT-Coder-V2.5-Dev | 27.95 GiB | 6.927 | — |
| North-Mini-Code-1.0 | 23.76 GiB | 6.696 | — |
| MiniMax-M2.7 | 175.15 GiB | 6.578 | — |
| Nanbeige4.2-3B | 3.09 GiB | 6.366 | — |
| MiniMax-M3 | 330.08 GiB | 6.639 | — |
| NVIDIA-Nemotron-3-Super-120B-A12B-BF16 | 106.87 GiB | 7.427 | — |
| Mistral-Small-4-119B-2603 | 92.60 GiB | 6.662 | — |
| Huihui-gemma-4-26B-A4B-it-abliterated | 21.33 GiB | 6.903 | — |
| MiMo-V2.5 | 239.34 GiB | 6.615 | — |
| GLM-5.1 | 578.66 GiB | 6.594 | — |
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | 429.09 GiB | 6.576 | — |
| MiMo-V2.5-Pro | 788.42 GiB | 6.619 | — |
●From the filewhat these mean