TensorSharp Adds Support for Qwen Image Editing
fuzhongkai · reddit · 2026-07-15
TensorSharp has updated its support for Unsloth Qwen Image Edit 2511, allowing users to run local image editing/outfit-changing demos using a combination of multiple components.
The post lists the required models: Qwen-Image-Edit's MMDiT DiT, Qwen Image VAE, Qwen2.5-VL-7B text encoder, and an optional Lightning LoRA. It also provides the startup command for TensorSharp.Server, which includes an OpenAI/Ollama compatible API and a WebUX chat interface.
Regarding performance, compared to stable-diffusion.cpp under the Qwen-Image-Edit 2511 + Lightning 4-step LoRA scenario:
- Warm total: 40.44s vs 48.16s, making TensorSharp 1.19x faster
- Per step: 7.57s vs 9.43s, 1.25x faster
- Sampling: 30.27s vs 37.73s, 1.25x faster
- However, text encoding is slower, while VAE encode/decode is faster
The author adds that TensorSharp also achieves performance close to llama.cpp when running autoregressive LLMs.
More from Infra
- SkyPilot exits stealth with $20M seed round and an AI compute platform for fragmented clouds — jfiance · 2026-07-22
- Nothing phone mockup turns a film joke into a modular design meme — ZeYanjie · 2026-07-22
- Actual Computer says its inference stack is tuned for Nvidia’s consumer Blackwell lineup — markjeffrey · 2026-07-22
- Ben Bajarin says CPU demand is still being badly underestimated — BenBajarin · 2026-07-22
- An energy model says the U.S. could run short of natural gas starting in 2028 — churchkey · 2026-07-22
- Devin adds e2b sandboxes for remote agent execution — badphilosopher · 2026-07-22