SDXL Turbo ConvRot (INT8, INT6 & INT4) for ComfyUI

ConvRot-quantized single-file ComfyUI checkpoints of stabilityai/sdxl-turbo in three clean bit widths: Full INT8, Full INT6 (w6a8_int8), and Full INT4 (asym_w4a8_int8).

Built with ConvRot (block Hadamard rotation) + Act-Order GPTQ Hessian error compensation + Hessian-weighted per-row clip search (INT8) / 2-pass through-grid FP8 group-scale search, damped coordinate-descent code refinement & Lloyd-Max codebooks + calibration bias correction, keeping Conv2d layers, text encoders (CLIP-L / OpenCLIP-G), and VAE unquantized in FP16.

โš ๏ธ License Separation Notice: This model is hosted in a dedicated repository because it is governed by the Stability AI Community License Agreement rather than Open RAIL++-M or Apache-2.0. Please review the license terms below before commercial or downstream use.


Available Checkpoints

Precision Tier File (ComfyUI/models/checkpoints/) Size (vs 6.94 GB FP16) Quantized UNet Linears
Full INT8 ConvRot sdxl-turbo-1.0-int8-convrot.safetensors 4.72 GB (-32.0%) 739 / 739 INT8
Full INT6 ConvRot sdxl-turbo-1.0-int6-convrot.safetensors 4.31 GB (-37.8%) 658 / 739 INT6 (W6A8) + 81 INT8 (K=640/det)
Full INT4 ConvRot sdxl-turbo-1.0-int4-convrot.safetensors 3.77 GB (-45.6%) 658 / 739 INT4 (W4A8) + 81 INT8 (K=640/det)

All files load directly in stock ComfyUI (Load Checkpoint, ComfyUI >= v0.45.0) using native comfy-kitchen CUDA kernels โ€” no extra nodes or unpacking required. (For SDXL Base/Lightning/Turbo, 80 small K=640 linear layers plus the single architecture-detection to_k key remain in ConvRot INT8 because comfy-kitchen w6a8/w4a8 kernels require K % 256 == 0.)


Visual Comparisons (2 steps, euler_ancestral, 512x512)

Turbo Comparison p0 Turbo Comparison p3 Turbo Comparison p7


Evaluation Metrics (24 renders: 8 prompts x 3 seeds, 512x512, ComfyUI v0.45.3)

Measured via scripts/verify_sdxl_quant.py directly against unquantized FP16 (sd_xl_turbo_1.0_fp16.safetensors):

  • Unquantized BF16 Reference (vs FP16): Cosine 0.99812 (min 0.99472), SSIM 0.92574 (min 0.87324), PSNR 31.22 dB (min 27.34 dB)
Precision Tier Cosine Mean / Min SSIM Mean / Min PSNR Mean / Min Warm Speed/Img Report
Full INT8 ConvRot 0.99890 / 0.99672 0.97214 / 0.93937 35.18 / 30.80 dB 0.19s g4_turbo_int8_summary.json
Full INT6 ConvRot 0.99777 / 0.99261 0.95993 / 0.92401 33.03 / 28.37 dB 0.19s g4_turbo_int6_summary.json
Full INT4 ConvRot 0.98796 / 0.92047 0.90092 / 0.66080 28.11 / 20.40 dB 0.20s g4_turbo_int4_summary.json

How to Use in ComfyUI

  1. Download INT8, INT6, or INT4 .safetensors into ComfyUI/models/checkpoints/.
  2. Load directly with ComfyUI's built-in Load Checkpoint node (ComfyUI >= v0.45.0).

Stability AI Community License & Acceptable Use Policy

This model is a Derivative Work of stabilityai/sdxl-turbo and is distributed under the Stability AI Community License Agreement (sai-nc-community), Copyright ยฉ Stability AI Ltd. All Rights Reserved. See also the included NOTICE file.

Key terms under LICENSE.md:

Downloads last month
62
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for SUSSYEATING/SDXL-Turbo-ConvRot

Quantized
(10)
this model