Qwen-Image-2.1-Turbo: From Open Release to Hands-On Tests

Episode 1 · Qwen Open-Sources Qwen-Image-2.1-Turbo: 2K Image Generation in 8 Denoising Steps (2026-10-09, 17 posts)

Alibaba's Tongyi Qwen team released and open-sourced Qwen-Image-2.1-Turbo on October 9. It is a sped-up version of Qwen-Image-2.1, with weights available on Hugging Face under Qwen/Qwen-Image-2.1-Turbo and the API launched alongside it. The model dramatically improves generation speed while retaining the original model's capabilities, marking another significant update for open-source image generation.

Confirmed

• The model is an accelerated checkpoint of Qwen-Image-2.1, using accelerated distillation to compress denoising from dozens of steps down to 8 steps, with no loss in quality per the team
• Retains the same 7B vision generation architecture as the original, supporting text-to-image and image editing
• Still generates 2K resolution images and supports image editing via natural language instructions
• Defaults to CFG=1 and adds KV caching for further speed gains
• Available through both diffusers and official usage paths, with weights open for download and the API live simultaneously

Why it matters

• 8-step denoising sharply cuts inference cost and wait times, bringing high-quality 2K image generation closer to practical, even real-time, use cases
• Fully open-source (downloadable weights) plus an API channel serve both local deployment and cloud calling needs
• In competition with closed-source image generation services, it gives the community an open-source option balancing speed and quality

• Qwen releases Qwen-Image-2.1-Turbo text-to-image model on Hugging Face — Qwen · 2026-10-09
• Alibaba open-sources Qwen-Image-2.1-Turbo, generating 2K images in just 8 denoising steps — Alibaba_Qwen · 2026-10-09
• Qwen-Image-2.1-Turbo Lands on Hugging Face With 8-Step Sampling, CFG=1 and Prefix KV Caching — SysPsych · 2026-10-09
• Qwen Releases Qwen-Image-2.1-Turbo: Open-Weights Image Model That Needs Only 8 Denoising Steps — ResearchCrafty1804 · 2026-10-09
• Alibaba's Qwen releases Qwen-Image-2.1-Turbo on Hugging Face — chocofoxy · 2026-10-09
• Qwen-Image-2.1-Turbo officially released on Hugging Face — CeFurkan · 2026-10-09
• Qwen-Image-2.1-Turbo officially released, weights live on Hugging Face — CeFurkan · 2026-10-09
• Qwen releases 8-step accelerated Qwen-Image-2.1-Turbo on Hugging Face — multimodalart · 2026-10-09
• Alibaba open-sources Qwen-Image-2.1-Turbo: 2K images in just 8 denoising steps — CurieuxExplorer · 2026-10-09
• Qwen Image 2.1 Turbo Lands on Hugging Face: 8 Steps, Comparable Quality — multimodalart · 2026-10-09
• Qwen Image 2.1 Turbo lands on Hugging Face: 8 steps, same quality — multimodalart · 2026-10-09
• Qwen-Image-2.1-Turbo released: 7B open image model in 8 denoising steps — lmoroney · 2026-10-10
• Qwen-Image-2.1-Turbo cuts 7B image model to 8 denoising steps — lmoroney · 2026-10-10
• Qwen-Image-2.1 ships official 8-step Turbo model, fastest inference stack still open — Xianbao_QIAN · 2026-10-10
• Qwen-Image-2.1-Turbo cuts image generation and editing to just 8 denoising steps — bdsqlsz · 2026-10-10
• Qwen-Image-2.1-Turbo released: 7B model generates 2K images in just 8 denoising steps — aziz4ai · 2026-10-10
• Alibaba open-sources Qwen-Image-2.1-Turbo: 8-step denoising for 2K image generation — pbaylies · 2026-10-11

Episode 2 · Qwen-Image-2.1-Turbo Gains Official ComfyUI Support (2026-10-09, 2 posts)

Comfy-Org has released official ComfyUI support for Qwen-Image-2.1-Turbo on Hugging Face, allowing users to run the Qwen image generation model directly in ComfyUI workflows.

• Qwen-Image-2.1-Turbo gets official ComfyUI support on Hugging Face — Time-Teaching1926 · 2026-10-09
• Qwen-Image-2.1-Turbo ships with official ComfyUI support on Hugging Face — Time-Teaching1926 · 2026-10-10

Episode 3 · Qwen Image 2.1 int8 quantized runs locally: turbo generates in 3 seconds with 8 steps (2026-10-10, 3 posts)

Users released and tested int8-quantized weights for Qwen Image 2.1 and its turbo variant, running locally with 8-step generation in about 3 seconds. Side-by-side comparisons also examined different quantized turbo configurations with Kijai's turbo LoRA.

• Qwen Image 2.1 int8 quants: 25 steps in 12s, turbo 8 steps in 3s in ComfyUI — Ant_6431 · 2026-10-10
• Qwen Image 2.1 Turbo LoRA vs Turbo: Same 8 Steps, Side-by-Side Comparison — Ant_6431 · 2026-10-10
• Qwen Image 2.1 Turbo Runs Locally: int8 Quantized Config Delivers Images in 8 Steps — Ant_6431 · 2026-10-10