Quantization format
IQ2_S
IQ2_S has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 452 files measured
Nominal bpw
—
from the block layout
Measured average
2.712
452 files
Range
2.145–9.233
varies by architecture
File sizes
0.02 GiB+
up to 311.72 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 11.09 GiB | 2.650 | — |
| gemma-4-26B-A4B-it | 9.53 GiB | 3.083 | — |
| gemma-4-12B-it | 4.39 GiB | 3.150 | — |
| Hy3 | 86.43 GiB | 2.485 | — |
| gemma-4-31B-it | 11.25 GiB | 3.089 | — |
| Ornith-1.0-35B | 10.25 GiB | 2.541 | — |
| ThinkingCap-Qwen3.6-27B | 9.59 GiB | 3.011 | — |
| Qwen3.5-122B-A10B | 37.85 GiB | 2.599 | — |
| Qwen3-Coder-Next | 21.76 GiB | 2.346 | — |
| Qwen2.5-32B-Instruct | 9.67 GiB | 2.536 | — |
| Qwen3.5-35B-A3B | 11.09 GiB | 2.650 | — |
| Qwen3-30B-A3B | 8.59 GiB | 2.416 | — |
| Qwen2.5-Coder-32B-Instruct | 9.67 GiB | 2.536 | — |
| Qwen3.5-27B | 9.79 GiB | 3.026 | — |
| Laguna-S-2.1 | 33.77 GiB | 2.467 | — |
| Qwen2.5-Coder-14B-Instruct | 4.66 GiB | 2.710 | — |
| Qwen3-30B-A3B-Instruct-2507 | 8.14 GiB | 2.291 | — |
| Voxtral-Small-24B-2507 | 6.97 GiB | 2.466 | — |
| Llama-3.3-70B-Instruct | 20.71 GiB | 2.522 | — |
| DeepSeek-R1-Distill-Llama-70B | 20.71 GiB | 2.522 | — |
| Step-3.7-Flash | 57.93 GiB | 2.471 | — |
| gemma-3-27b-it | 8.18 GiB | 2.561 | — |
| GLM-4.6V-Flash | 3.52 GiB | 2.941 | — |
| DeepSeek-R1-Distill-Qwen-14B | 4.66 GiB | 2.710 | — |
| Mistral-Small-24B-Instruct-2501 | 6.96 GiB | 2.538 | — |
●From the filewhat these mean