Quantization format
UD_IQ2_XXS
UD_IQ2_XXS has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 199 files measured
Nominal bpw
—
from the block layout
Measured average
2.591
199 files
Range
1.596–5.374
varies by architecture
File sizes
0.17 GiB+
up to 312.21 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-27B | 8.74 GiB | 2.704 | — |
| Qwen3.6-35B-A3B | 10.02 GiB | 2.394 | — |
| Qwen3.5-9B | 2.97 GiB | 2.644 | — |
| gemma-4-26B-A4B-it | 9.24 GiB | 2.990 | — |
| DeepSeek-V4-Flash | 84.62 GiB | 2.498 | — |
| Qwen3.5-4B | 1.42 GiB | 2.610 | — |
| Qwen3-Coder-30B-A3B-Instruct | 9.62 GiB | 2.708 | — |
| gemma-4-31B-it | 7.95 GiB | 2.183 | — |
| Qwen3-VL-30B-A3B-Instruct | 9.63 GiB | 2.663 | — |
| Qwen-AgentWorld-35B-A3B | 10.71 GiB | 2.654 | — |
| Llama-3.2-1B-Instruct | 0.43 GiB | 3.006 | — |
| Qwen3-8B | 2.43 GiB | 2.545 | — |
| Qwen3.5-0.8B | 0.31 GiB | 3.098 | — |
| Qwen3-4B | 1.17 GiB | 2.497 | — |
| GLM-5.2 | 222.08 GiB | 2.532 | — |
| Llama-3.1-8B-Instruct | 2.33 GiB | 2.495 | — |
| Ornith-1.0-35B | 10.71 GiB | 2.654 | — |
| Llama-3.2-3B-Instruct | 0.97 GiB | 2.606 | — |
| Kimi-K2.7-Code | 296.00 GiB | 2.402 | — |
| Qwen3.5-122B-A10B | 34.12 GiB | 2.343 | — |
| Qwen3-Coder-Next | 21.71 GiB | 2.341 | — |
| gemma-3-1b-it | 0.53 GiB | 4.516 | — |
| Qwen3.5-35B-A3B | 9.93 GiB | 2.371 | — |
| Qwen3-14B | 4.16 GiB | 2.421 | — |
| Qwen3-VL-4B-Instruct | 1.17 GiB | 2.263 | — |
●From the filewhat these mean