Tencent's new Hy Image3.5 image generation model offers a 30% higher win rate in human evaluations than the previous 3.0 version. The preview version supports both text-to-image and image-to-image generation with a maximum output resolution of 2K. Access via the Tencent Cloud API costs $0.024 per image, charging users only for generated content rather than the reference images provided.
An internal blind test involving hundreds of designers placed the model on par with Seedream 5.0 Pro and slightly ahead of Qwen-Image-3.0 Pro and Nano Banana Pro. The tool allows for up to 5 reference images per generation and supports multi-turn editing with conversation context. Two week free trials are currently available on the OnSolo and Miora platforms.
Key sources
- SOURCE@tencenthunyuan“Professional-grade image generation, +30% win rate in human eval vs Hy Image3.0”x.com
- SUPPORT@chinazhidx“In Tencent’s internal GSB blind test with hundreds of designers: • ~30% higher than Hy-Image 3.0 • On par with Seedream 5.0 Pro • Slightly ahead of Nano Banana Pro and Qwen-Image-3.0 Pro”x.com
- SOURCEmarketbrief.now