Breaking

Tencent Rolls Out Hy-Image-3.0 as Alibaba Opens Its Apsara AI Conference

September 21, 2026 at 23:37 EDT

Media Generation
Foundation Models

Tencent has launched Hy-Image-3.0, its newest image-generation model, sharpening China’s generative-AI rivalry on the same day Alibaba opened its 2026 Apsara Conference in Hangzhou. The release gives Tencent a fresh visual-generation entry as Alibaba, ByteDance and other Chinese AI groups compete for developers, cloud customers and creative-tool adoption.

September 22, 2026 · Tencent

Tencent Enters The Week With Hy-Image-3.0

Tencent’s newest Hunyuan image model arrives as Alibaba opens Apsara, turning China’s image-generation race into a contest of model quality, cloud access and developer usability.

8,192

character prompt limit

preset image sizes

80B+

total parameters

13B

active parameters per token

Model Scale, Drawn To Size

13B

active per token

80B+

total parameters

The system is large, but only a subset of parameters is active for each inference step.

What developers like

Open availability, strong prompt-following ambitions and emerging local workflow support make the HunyuanImage line appealing to experimenters.

Where friction remains

Large model size, memory pressure and workflow integration gaps can slow adoption in everyday creator pipelines.

The competitive frame

Cloud API access→Image generation→Creator workflows→China AI rivalry

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.