Stuubzzz · 视频工作流实测·· 2026-10-18PicksAI score66
Hands-on test of Qwen-Image 2.1 Turbo in ComfyUI: 8-step scheduling and 300 renders
Qwen-Image 2.1 Turbo in ComfyUI: Qwen's 8-step schedule, settings and speed in 300 renders
AI briefing
The author tested Qwen-Image 2.1 Turbo with int8 files on an RTX 5090: native 2K (4 MP) images averaged 11.8 seconds, versus 36.9 seconds for the 30-step base model, about 3.1 times faster.
Why it matters
The author used 300 renders to test the differences between Turbo and the base model, and provides reusable 8-step sigma and resolution settings.
Open original stuubzzz.studio
Image GenerationComfyUI & WorkflowsTutorials & Practice#Tutorial#Image generation#ComfyUI & workflows#Benchmarks#Alibaba