Tencent has introduced the Hy Image3.5 preview model, designed for professional image generation. The tool supports text-to-image, image-to-image, and multi-turn conversational creation. It allows users to upload up to five reference images and generates outputs in various aspect ratios with a maximum resolution of 2K.
Internal testing involving hundreds of professional designers showed a 30% performance improvement over the Hy Image3.0 model. The new version is now available through Tencent Cloud TokenHub, with pricing set at $0.024 per image for international users and ¥0.15 per 2K image for domestic clients.
The model has been integrated into several applications, including Yuanbao, WorkRally, OnSolo, Miora, WorkBuddy, and ima. It is also connected to Tencent Cloud Media Processing and Video on Demand platforms. Some applications, such as Miora and WorkRally, currently offer a two-week free trial period.
Tencent has been testing the model in the Yuanbao app for two months, reporting a 50% increase in total image generation requests. The company maintains a development cycle where major model versions are updated every two months, utilizing a preview-first approach to incorporate user feedback into the final releases.
