Tencent launches Hy Image 3.5 preview: +30% human eval win rate, $0.024 per image
TencentHunyuan · x · 2026-09-23
Tencent Hunyuan has launched Hy Image 3.5 preview, a professional-grade image generation and editing model:
- +30% win rate in human eval vs Hy Image 3.0
- Supports both Text to Image and Image to Image, up to 2K resolution with better consistency
- Uses deep reasoning before editing/creating: works without detailed prompts, follows instructions reliably, and handles composition well. Reviewers note it pushes old-photo restoration to a new level without destroying faces or human features
- API priced at $0.024 per image on Tencent Cloud, with reference images free of charge; two-week free trials available on Miora and OnSolo
Related event: Tencent Hunyuan Launches Hy Image 3.5 Preview on ComfyUI(2 posts)→
More from Multimodal
- Spirula Studio: cross-vendor 3D Gaussian Splatting trainer turns video into mesh — harry7557558 · 2026-09-23
- Ming-Image-0.1-Design ships open weights with native RGBA output and layer separation — ActNo1693 · 2026-09-23
- Ming-Image-0.1-Design released: 6B design generation plus a 6B RGBA layer-splitter — harshu24118 · 2026-09-23
- Dev builds free Colab Gradio UI for Z-Image Turbo with upscaling and face fix — OkEnvironment1732 · 2026-09-23
- Same prompt, two new models: Claude Opus 5.5 vs GPT-6 Sol launch video comparison — CurieuxExplorer · 2026-09-23
- Midjourney demo: mechanical hummingbird with full sref-stacking parameter recipe — ciguleva · 2026-09-23