Quantization format
Q3_K
Q3_K has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 84 files measured
Nominal bpw
—
from the block layout
Measured average
4.493
84 files
Range
3.553–14.091
varies by architecture
File sizes
0.02 GiB+
up to 104.93 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| whisper-medium | 0.32 GiB | 3.601 | — |
| whisper-large-v3 | 0.64 GiB | 3.553 | — |
| whisper-large-v3-turbo | 0.34 GiB | 3.636 | — |
| Z-Image-Turbo | 2.93 GiB | 4.086 | — |
| Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled | 15.99 GiB | 3.820 | — |
| Wan2.2-Animate-14B | 8.14 GiB | 4.048 | — |
| FLUX.1-Fill-dev | 4.99 GiB | 3.601 | — |
| Llama-3.1-8B | 3.74 GiB | 4.004 | — |
| WAN2.2-14B-Rapid-AllInOne | 8.03 GiB | 3.979 | — |
| Codestral-22B-v0.1 | 10.02 GiB | 3.868 | — |
| whisper-small | 0.11 GiB | 3.768 | — |
| DeepSeek-Coder-V2-Lite-Instruct | 7.57 GiB | 4.139 | — |
| Hermes-3-Llama-3.1-8B | 3.74 GiB | 4.004 | — |
| Qwen2-0.5B-Instruct | 0.33 GiB | 5.756 | — |
| Qwen2.5-Math-7B-Instruct | 3.55 GiB | 4.001 | — |
| gemma-2-27b-it | 12.50 GiB | 3.945 | — |
| MiniCPM-V-2_6 | 3.55 GiB | 3.760 | — |
| s2-pro | 2.82 GiB | 5.312 | — |
| whisper-base | 0.03 GiB | 4.088 | — |
| glm-4-9b-chat | 4.72 GiB | 4.310 | — |
| G4-Alice-v1.2-31B | 14.24 GiB | 3.911 | — |
| Qwen2.5-Math-1.5B-Instruct | 0.77 GiB | 4.271 | — |
| Meta-Llama-3.1-8B-Instruct-abliterated | 3.74 GiB | 4.004 | — |
| Phi-3-mini-4k-instruct | 1.82 GiB | 4.094 | — |
| openchat-3.6-8b-20240522 | 3.74 GiB | 4.004 | — |
●From the filewhat these mean