Quantization format
Q5_K
Q5_K has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 108 files measured
Nominal bpw
—
from the block layout
Measured average
6.374
108 files
Range
4.174–21.264
varies by architecture
File sizes
0.03 GiB+
up to 208.25 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| embeddinggemma-300m | 0.30 GiB | 8.417 | — |
| Qwythos-9B-Claude-Mythos-5-1M | 6.19 GiB | 5.647 | — |
| FLUX.2-klein-9B | 6.35 GiB | 6.004 | — |
| whisper-medium | 0.50 GiB | 5.647 | — |
| whisper-large-v3 | 1.01 GiB | 5.604 | — |
| whisper-large-v3-turbo | 0.53 GiB | 5.677 | — |
| Qwen3-ASR-1.7B | 1.63 GiB | 5.942 | — |
| Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled | 23.61 GiB | 5.640 | — |
| Wan2.2-Animate-14B | 12.21 GiB | 6.074 | — |
| Mistral-Small-24B-Instruct-2501 | 15.61 GiB | 5.689 | — |
| Mistral-7B-Instruct-v0.3 | 4.78 GiB | 5.669 | — |
| Llama-3.1-8B | 5.34 GiB | 5.711 | — |
| WAN2.2-14B-Rapid-AllInOne | 12.13 GiB | 6.010 | — |
| mistral-small-3.1-24b-instruct-2503-hf | 15.61 GiB | 5.689 | — |
| gemma-2-9b-it | 6.19 GiB | 5.754 | — |
| Qwen3.6-27B-DFlash | 1.14 GiB | 5.668 | — |
| Codestral-22B-v0.1 | 14.64 GiB | 5.654 | — |
| parakeet-ctc-0.6b | 8.11 GiB | — | — |
| whisper-small | 0.16 GiB | 5.798 | — |
| DeepSeek-Coder-V2-Lite-Instruct | 11.04 GiB | 6.036 | — |
| Gemma-4-31B-StyleTune | 18.17 GiB | 4.777 | — |
| Qwen3.6-35B-A3B-DFlash | 0.26 GiB | 5.810 | — |
| DA3-BASE | 1.35 GiB | — | — |
| Hermes-3-Llama-3.1-8B | 5.34 GiB | 5.711 | — |
| Qwen2-0.5B-Instruct | 0.39 GiB | 6.803 | — |
●From the filewhat these mean