QuantFunc A4W4 INT4 — quantized from Qwen-Image-2.1

True 4-bit inference — A4W4 (4-bit activations × 4-bit weights).

QuantFunc's core quantized matrix multiplications use INT4 activations and INT4 weights on its INT4 inference backend. A4W4 describes the compute precision used during inference, alongside the reduced weight size and memory bandwidth demand.

QuantFunc INT4 compresses Qwen-Image-2.1 to about a quarter of its 16-bit size. In our visual comparisons, composition, text rendering, detail and style stay close to the 16-bit baseline.

Showcase

All images below were generated by Qwen-Image-2.1-QuantFunc-4bit.

Floral ad design Portrait photography Chinese food poster Whimsical everyday scene

Swap one loader, keep your workflow

1. Install or update ComfyUI-QuantFunc.
2. Download the r128 or r32 weights.
3. Swap your model loader for the QuantFunc loader and pick the matching weight file.

Every other node, connection and generation parameter stays as-is.

RTX 20-series through GB300, one build covers it all

Runs on every NVIDIA SM75+ GPU: RTX 20/30/40/50-series, A100, H100, H200, B100, B200, GB300.

4-bit weights significantly cut the weight-bandwidth cost of loading and inference, making Qwen-Image-2.1 much easier to run on consumer GPUs. Actual VRAM needs depend on resolution, batch size and the rest of your workflow.

Choose a model

Variant File Size Use case r128 qwen_image_21_base_quantfunc_int4_r128.safetensors 4.66 GB (4.34 GiB) Quality-first, recommended r32 qwen_image_21_base_quantfunc_int4_r32.safetensors 4.33 GB (4.03 GiB) Size/VRAM-first

Loading

Weights use QuantFunc's own safetensors format — load with ComfyUI-QuantFunc or the QuantFunc inference engine, not as a drop-in Diffusers checkpoint.

If the model isn't recognized or fails to load, update ComfyUI-QuantFunc to the latest version and restart ComfyUI.

Source & license

This repository distributes derived quantized weights produced from Qwen/Qwen-Image-2.1, released by the Qwen team.

Qwen-Image-2.1 and its derivative weights follow the Qwen Research License (license text). QuantFunc's quantization tooling and implementation follow their own respective licenses. Please read and comply with the original model's license terms before use.

Official QuantFunc links

• Hugging Face — QuantFunc
• ModelScope — QuantFunc
• Official Discord
• QuantFunc website
• ComfyUI-QuantFunc plugin and workflows

License terms

Built with Qwen. These derived weights retain the Qwen Research License: the community grant is for research or evaluation purposes only. Commercial use requires a separate license from the original licensor. Redistribution must include the original agreement, notices of modification, and the required Notice file. See the full license.

Source and weight integrity

Original QuantFunc release: QuantFunc/Qwen-Image-2.1-4bit.

• qwen_image_21_base_quantfunc_int4_r128.safetensors — SHA256: 920a3be0b5b559ea67e97586f68fae72aa0fd60ab97537ebdcfbf1a598634776
• qwen_image_21_base_quantfunc_int4_r32.safetensors — SHA256: 177a9b59a7d65ad41925cb3fc29d9a8a7b1deb5732d7aea1b36405f35b612653

Showcase media is reproduced from the original QuantFunc model card. Exact seeds, prompts and rank variants are not supplied for every example. Performance figures are QuantFunc's reported measurements under the stated conditions; they are not guarantees for other workflows.