Quantization format

I1_IQ3_S

I1_IQ3_S has a nominal rate of — bits per weight. We have no measured files carrying this label yet.

From the file· 1036 files measured

Nominal bpw

—

from the block layout

Measured average

3.698

1036 files

Range

2.045–7.900

varies by architecture

File sizes

0.08 GiB+

up to 159.92 GiB

Real files

one per model, most downloaded first

Model	Size●	Effective bpw●	vs nominal●
Qwopus3.6-35B-A3B-v1	14.20 GiB	3.393	—
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP	11.74 GiB	3.629	—
Qwen3-VL-8B-Instruct-abliterated-v1	3.53 GiB	3.458	—
Qwythos-9B-v2	4.17 GiB	3.709	—
gemma-4-26B-A4B-it-ultra-uncensored-heretic	11.38 GiB	3.789	—
Sugoi-14B-Ultra-HF	6.23 GiB	3.625	—
gemma-4-12b-heretic-abliterated	5.15 GiB	3.698	—
FastContext-1.0-4B-SFT	1.77 GiB	3.778	—
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved	11.74 GiB	3.685	—
gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking	12.82 GiB	3.520	—
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled	14.20 GiB	3.393	—
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning	3.43 GiB	3.668	—
Huihui-ThinkingCap-Qwen3.6-27B-abliterated	11.74 GiB	3.685	—
Mistral-7B-Instruct-v0.2	2.96 GiB	3.516	—
Huihui-Qwen3-Coder-Next-abliterated	32.18 GiB	3.469	—
DeepSeek-R1-Distill-Llama-70B-abliterated	28.79 GiB	3.505	—
gemma-4-26B-A4B-it-heretic	11.38 GiB	3.789	—
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-uncensored-heretic	5.15 GiB	3.698	—
Qwen3.5-122B-A10B-heretic-v2	49.29 GiB	3.467	—
Llama-3.3-70B-Instruct-abliterated	28.79 GiB	3.505	—
Qwable-9B-Claude-Fable-5	4.07 GiB	3.716	—
EVA-Qwen2.5-14B-v0.2	6.23 GiB	3.625	—
Prosopon-31B	13.38 GiB	3.517	—
Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW	4.04 GiB	3.755	—
Qwen3.6-34B-80L-Fable-5-Heretic	14.26 GiB	3.662	—

●From the filewhat these mean