Qwen released Qwen-Image-2.1-Turbo weights with eight-step sampling and Diffusers usage in the model card.
Qwen releases an eight-step turbo image model
2026-10-09Verified announcementTRIAL
Qwen released Qwen-Image-2.1-Turbo weights. The 7B visual-generation model handles text-to-image and editing in eight denoising steps, uses the existing QwenImage21Pipeline and supports outputs up to the 2K class. The model card provides no comparative benchmark against the base model.
What this means for you
For teams needing local image generation, eight-step sampling may cut latency materially, but speed, memory, text rendering and edit fidelity need direct testing. Review the custom license and dependency on unreleased Diffusers source first.
Evidence & related material
Technical details & scores
| Parameters | 7B visual-generation architecture |
|---|---|
| Sampling | 8 denoising steps; CFG 1 by default |
| Tasks | text-to-image and image editing |
| Resolution | up to 2K-class output |
| Runtime | QwenImage21Pipeline; requires latest Diffusers source at release |
| License | custom qwen-research license |
| Benchmark | no comparative benchmark in model card |
Technical impact 3
Engineering value 4
Adoption signal 2
Maturity 3
Verification cost 2
Risk 3