Quantization format
Q6_K_M
Q6_K_M has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 10 files measured
Nominal bpw
—
from the block layout
Measured average
6.415
10 files
Range
3.644–7.657
varies by architecture
File sizes
0.41 GiB+
up to 24.76 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| GRM-2.6-Plus-0628 | 24.76 GiB | 7.657 | — |
| LFM2-1.2B | 0.90 GiB | 6.582 | — |
| gemma-4-12B | 9.34 GiB | 6.709 | — |
| LFM2-2.6B | 1.97 GiB | 6.572 | — |
| Ministral-3-8B-Instruct-2512-BF16 | 6.98 GiB | 6.720 | — |
| granite-3.1-8b-instruct | 6.24 GiB | 6.565 | — |
| PaddleOCR-VL-1.5 | 0.41 GiB | 3.644 | — |
| Magistral-Small-2509-Vision | 18.32 GiB | 6.554 | — |
| orpheus-3b-0.1-pretrained | 2.90 GiB | 6.582 | — |
| granite-3.1-2b-instruct | 1.94 GiB | 6.570 | — |
●From the filewhat these mean