Tencent is developing AI “world models” that simulate physical environments, entering a field also pursued by Google DeepMind and xAI.

The company has open-sourced HunyuanWorld 1.0, HunyuanWorld-Voyager, and FlashWorld, which let developers create interactive 3D worlds from images or text.

Tencent has integrated these models into its robotics training platform and gaming business to improve virtual-world generation.

According to Guo Chunchao, these models allow AI agents and robots to learn tasks in simulations, reducing real-world risks.

World models are seen as essential for advancing spatial intelligence, though concerns remain about how reliable simulations are for real-world robotics.

Google DeepMind’s Genie 3 and other recent launches show rising global interest in this technology.

🔗 Source: South China Morning Post

Recent Tencent developments