Quantization format

I1_IQ2_XXS

I1_IQ2_XXS has a nominal rate of — bits per weight. We have no measured files carrying this label yet.

From the file· 957 files measured
Nominal bpw
—
from the block layout
Measured average
2.445
957 files
Range
1.483–5.207
varies by architecture
File sizes
0.06 GiB+
up to 96.39 GiB

Real files

one per model, most downloaded first
ModelSize●Effective bpw●vs nominal●
Qwopus3.6-35B-A3B-v18.85 GiB2.113—
Qwen3-VL-8B-Instruct-abliterated-v12.32 GiB2.272—
gemma-4-26B-A4B-it-ultra-uncensored-heretic8.66 GiB2.883—
Sugoi-14B-Ultra-HF4.02 GiB2.336—
gemma-4-12b-heretic-abliterated3.32 GiB2.387—
FastContext-1.0-4B-SFT1.16 GiB2.479—
gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking8.08 GiB2.219—
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled8.85 GiB2.113—
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning2.23 GiB2.390—
Mistral-7B-Instruct-v0.21.85 GiB2.200—
Huihui-Qwen3-Coder-Next-abliterated19.57 GiB2.110—
DeepSeek-R1-Distill-Llama-70B-abliterated17.79 GiB2.165—
gemma-4-26B-A4B-it-heretic8.66 GiB2.883—
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-uncensored-heretic3.32 GiB2.387—
Qwen3.5-122B-A10B-heretic-v229.99 GiB2.110—
Llama-3.3-70B-Instruct-abliterated17.79 GiB2.165—
Qwable-9B-Claude-Fable-52.89 GiB2.637—
EVA-Qwen2.5-14B-v0.24.02 GiB2.336—
Prosopon-31B8.51 GiB2.237—
Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW2.63 GiB2.445—
Gemma-4-26B-A4B-StyleTune-V28.89 GiB2.876—
Qwen2.5-VL-7B-Instruct-abliterated2.12 GiB2.193—
Qwen3.5-9B-Claude-4.6-Opus-Deckard-V4.2-Uncensored-Heretic-Thinking2.89 GiB2.637—
MythoMax-L2-13b3.50 GiB2.307—
Mistral-7B-Instruct-v0.11.85 GiB2.200—
●From the filewhat these mean

Other quantizations

Q2_KQ3_K_SQ3_K_MQ3_K_LQ4_0Q4_K_SQ4_K_MQ5_K_SQ5_K_MQ6_KQ8_0IQ4_XSIQ3_XXSIQ2_XXSF16BF16F32