Quantization format
UD_IQ3_S
UD_IQ3_S has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 32 files measured
Nominal bpw
—
from the block layout
Measured average
3.250
32 files
Range
2.903–4.560
varies by architecture
File sizes
3.33 GiB+
up to 390.04 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 12.74 GiB | 3.043 | — |
| gemma-4-26B-A4B-it | 10.51 GiB | 3.402 | — |
| DeepSeek-V4-Flash | 109.25 GiB | 3.226 | — |
| Qwen-AgentWorld-35B-A3B | 13.96 GiB | 3.458 | — |
| GLM-5.2 | 287.44 GiB | 3.278 | — |
| Ornith-1.0-35B | 13.96 GiB | 3.458 | — |
| Kimi-K2.7-Code | 390.04 GiB | 3.165 | — |
| Qwen3.5-122B-A10B | 43.36 GiB | 2.978 | — |
| Qwen3-Coder-Next | 27.65 GiB | 2.981 | — |
| Qwen3.5-35B-A3B | 12.65 GiB | 3.023 | — |
| Laguna-S-2.1 | 45.10 GiB | 3.296 | — |
| Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 | 17.53 GiB | 4.560 | — |
| Step-3.7-Flash | 74.54 GiB | 3.180 | — |
| LFM2.5-8B-A1B | 3.33 GiB | 3.374 | — |
| Ornith-1.0-397B | 150.77 GiB | 3.264 | — |
| Qwen3.5-397B-A17B | 136.32 GiB | 2.903 | — |
| North-Mini-Code-1.0 | 11.89 GiB | 3.350 | — |
| MiniMax-M2.7 | 77.87 GiB | 2.925 | — |
| MiniMax-M3 | 162.75 GiB | 3.274 | — |
| NVIDIA-Nemotron-3-Super-120B-A12B-BF16 | 52.74 GiB | 3.665 | — |
| Mistral-Small-4-119B-2603 | 41.36 GiB | 2.975 | — |
| Huihui-gemma-4-26B-A4B-it-abliterated | 10.45 GiB | 3.381 | — |
| MiMo-V2.5 | 106.98 GiB | 2.957 | — |
| GLM-5.1 | 260.39 GiB | 2.967 | — |
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | 233.30 GiB | 3.575 | — |
●From the filewhat these mean