Quantization format

I1_IQ4_XS

I1_IQ4_XS has a nominal rate of — bits per weight. We have no measured files carrying this label yet.

From the file· 1034 files measured

Nominal bpw

—

from the block layout

Measured average

4.413

1034 files

Range

1.080–9.445

varies by architecture

File sizes

0.10 GiB+

up to 153.13 GiB

Real files

one per model, most downloaded first

Model	Size●	Effective bpw●	vs nominal●
Qwopus3.6-35B-A3B-v1	17.44 GiB	4.168	—
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP	14.26 GiB	4.408	—
Qwen3-VL-8B-Instruct-abliterated-v1	4.25 GiB	4.163	—
Qwythos-9B-v2	4.96 GiB	4.414	—
gemma-4-26B-A4B-it-ultra-uncensored-heretic	12.96 GiB	4.315	—
Sugoi-14B-Ultra-HF	7.56 GiB	4.398	—
gemma-4-12b-heretic-abliterated	6.18 GiB	4.438	—
FastContext-1.0-4B-SFT	2.11 GiB	4.516	—
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved	14.26 GiB	4.477	—
gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking	15.59 GiB	4.281	—
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled	17.44 GiB	4.168	—
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning	4.14 GiB	4.431	—
Huihui-ThinkingCap-Qwen3.6-27B-abliterated	14.26 GiB	4.477	—
Mistral-7B-Instruct-v0.2	3.64 GiB	4.317	—
Huihui-Qwen3-Coder-Next-abliterated	39.68 GiB	4.278	—
DeepSeek-R1-Distill-Llama-70B-abliterated	35.30 GiB	4.298	—
gemma-4-26B-A4B-it-heretic	12.96 GiB	4.315	—
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-uncensored-heretic	6.18 GiB	4.438	—
Qwen3.5-122B-A10B-heretic-v2	60.76 GiB	4.274	—
Llama-3.3-70B-Instruct-abliterated	35.30 GiB	4.298	—
Qwable-9B-Claude-Fable-5	4.84 GiB	4.418	—
EVA-Qwen2.5-14B-v0.2	7.56 GiB	4.398	—
Prosopon-31B	16.28 GiB	4.280	—
Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW	4.83 GiB	4.487	—
Qwen3.6-34B-80L-Fable-5-Heretic	17.37 GiB	4.460	—

●From the filewhat these mean