Overview
Tencent Hunyuan released Hy Image 3.5 preview on September 22 as its high-value professional image model. It keeps the design approach of Hy Image 3.0 and targets three things: text rendering and layout inside the image, realism and style expression, and consistency across consecutive edits. A blind GSB test run by Tencent with several hundred professional designers put this version about 30% ahead of Hy Image 3.0, level with Seedream 5.0 pro and slightly ahead of Nano-Banana Pro and Qwen-Image-3.0 Pro. Those results come from Tencent’s internal testing.
Key Features
- Multi-turn creation: Edits build on previous context, so changing text, swapping elements or adjusting layout does not mean starting from scratch.
- Up to five reference images: Both text-to-image and image-to-image accept multiple references, so character, style and composition can be supplied at once.
- Up to 2K output: Supports multiple aspect ratios with output resolution up to 2K.
- Charged on output only: The Tencent Cloud API prices a 2K image at 0.15 CNY, and only the final output image is billed.
- Native in Tencent products: Already live in Yuanbao, ima, WorkBuddy, Miora, WorkRally and OnSolo, with the API on Tencent Cloud TokenHub, Media Processing Service and VOD.
Use Cases
- Commercial posters and brand visuals, where accurate text and stable layout are the direct target of this release
- Film and series production: building character, prop and scene asset images from a script at volume
- E-commerce assets, including product shots, virtual try-on and localized style changes
- Knowledge infographics that turn knowledge base content into illustrations, information graphics and slide backdrops
Pros
- At 0.15 CNY for a 2K image and billed only on output, batch production costs are easy to model
- Multi-turn editing puts the familiar problem of one edit changing the whole image at the center of this release
- During two months of staged rollout in Yuanbao, overall image generation requests rose more than 50%, with portrait retouching up 100%
- Available to developers and enterprises through Tencent Cloud TokenHub, Media Processing Service and VOD
- Hunyuan keeps a roughly two-month cadence for major versions, following the same preview-first, release-later pattern
Pricing
The Tencent Cloud API prices 2K images at 0.15 CNY each, or 0.024 USD on the international edition, charging only for output images. Developers and enterprises can call it through Tencent Cloud TokenHub. Yuanbao users get portrait editing, smart photo retouching, travel planning images and knowledge infographics for free, and Miora, WorkRally and OnSolo offer a two-week free trial. Check the official page for current prices and quotas.
Summary
Hy Image 3.5 preview is a routine iteration on the Hunyuan image line, aimed at what professional creators actually care about: whether text renders accurately, whether consecutive edits drift, and whether production at volume is affordable. At 0.15 CNY per output image with multi-turn editing, the cost of retries comes down, and Tencent’s internal testing puts it level with Seedream 5.0 pro. It is still a preview, so editing stability on complex layouts is something to watch as it ships.
Version History
- Tencent Hunyuan releases Hy Image 3.5 preview at 0.15 CNY per 2K image (2026-09-22): Tencent Hunyuan released Hy Image 3.5 preview with text-to-image, image-to-image and multi-turn creation, accepting up to five reference images and outputting up to 2K. The Tencent Cloud API prices 2K images at 0.15 CNY each and bills only for output images. An internal blind GSB test with several hundred designers showed a roughly 30% win-rate gain over Hy Image 3.0, on par with Seedream 5.0 pro. The model is live in Yuanbao, WorkRally, OnSolo, Miora, WorkBuddy and ima, with the API on Tencent Cloud TokenHub, Media Processing Service and VOD. During two months of staged rollout in Yuanbao, overall image generation requests rose more than 50%.