Quantization format
IQ1_M
IQ1_M has a nominal rate of — bits per weight. We have no measured files carrying this label yet.
From the file· 255 files measured
Nominal bpw
—
from the block layout
Measured average
2.170
255 files
Range
1.660–7.918
varies by architecture
File sizes
0.02 GiB+
up to 220.44 GiB
Real files
one per model, most downloaded first
| Model | Size● | Effective bpw● | vs nominal● |
|---|---|---|---|
| Qwen3.6-35B-A3B | 8.77 GiB | 2.096 | — |
| Hy3 | 66.22 GiB | 1.904 | — |
| gemma-4-31B-it | 9.42 GiB | 2.587 | — |
| Qwen3.5-122B-A10B | 29.61 GiB | 2.033 | — |
| Qwen3-Coder-Next | 16.11 GiB | 1.737 | — |
| Qwen3.5-35B-A3B | 8.77 GiB | 2.096 | — |
| jina-embeddings-v5-text-small | 0.20 GiB | 2.900 | — |
| Laguna-S-2.1 | 25.75 GiB | 1.881 | — |
| Phi-3.5-mini-instruct | 0.85 GiB | 1.920 | — |
| gemma-2-2b-it | 0.81 GiB | 2.674 | — |
| Inkling | 210.69 GiB | 1.900 | — |
| Llama-3.3-70B-Instruct | 15.60 GiB | 1.899 | — |
| DeepSeek-R1-Distill-Llama-70B | 15.60 GiB | 1.899 | — |
| Step-3.7-Flash | 44.41 GiB | 1.894 | — |
| Qwen2.5-72B-Instruct | 22.11 GiB | 2.612 | — |
| Mistral-7B-Instruct-v0.3 | 1.64 GiB | 1.940 | — |
| Meta-Llama-3-8B-Instruct | 2.01 GiB | 2.154 | — |
| Llama-3.1-70B-Instruct | 15.60 GiB | 1.899 | — |
| jina-embeddings-v5-text-nano | 0.09 GiB | 3.842 | — |
| Ornith-1.0-397B | 85.09 GiB | 1.842 | — |
| Qwen3.5-397B-A17B | 91.53 GiB | 1.949 | — |
| Qwen2-7B-Instruct | 1.90 GiB | 2.145 | — |
| Ornith-1.0-35B-AEON-Ultimate-Uncensored-BF16 | 7.67 GiB | 1.877 | — |
| Mixtral-8x22B-v0.1 | 30.48 GiB | 1.862 | — |
| DeepSeek-V3-0324 | 138.66 GiB | 1.740 | — |
●From the filewhat these mean