ComfyUI-Native-Int8-ConvRot

"INT8 ConvRot is row-wise INT8 with parameters and activations rotated before quantization via ConvRot." — ComfyUI-INT8-Fast Metrics.md

Models

Quality Ranking

Requirements

• ComfyUI >= v0.27.0. (Comfy-Org/ComfyUI#14636, merged June 2026). Update ComfyUI if your loader reports an invalid quantization type.
• Models load with the standard Load Diffusion Model node
• A pytorch build with cu130 support. If you are using an older version, you will encounter memory and performance issues.

How to Quantize a Model to INT8 ConvRot

1. Install silveroxides' convert_to_quant:
   pip install -U convert-to-quant

   PyTorch must be installed separately with the correct CUDA version.

2. Convert the model:
   bash

   1ctq -i source_model_bf16.safetensors -o converted_model_int8_convrot.safetensors \ 2 --int8 --scaling_mode row --simple --convrot --convrot-group-size [64,256,1024] \ 3 --comfy_quant --save-quant-metadata --<model-arch-flag>

Notes

• --convrot-group-size accepts 64, 256, or 1024. It is recommended to choose a value that divides evenly into all of the model's layer dimensions.
• --<model-arch-flag> selects the layer-exclusion preset for your model architecture (e.g. --wan, --flux2, --zimage). Run ctq --help-filters (or ctq -hf) for the full list.

References

1. Reddit: "So is INT8-ConvRot the new hot thing?"
2. ComfyUI-INT8-Fast — Metrics.md (benchmark methodology & full tables)
3. Comfy-Org/Boogu-Image discussion #10
4. ComfyUI PR #14636 — Support int8 models
5. bertbobson/ComfyUI-INT8_ConvRot
6. silveroxides/convert_to_quant

Licensing and Commercial Use