Alibaba Open-Sources 15B Video World Model AlayaWorld

Alibaba's AlayaLab team has officially open-sourced AlayaWorld, a full-stack video world model with 15B parameters. Designed to generate explorable and continuously evolving long-horizon virtual environments from text, images, or videos, AlayaWorld has garnered significant attention due to its high parameter count and open-source strategy.

Confirmed

AlayaWorld supports streaming video generation at 24 FPS with 540p and 720p resolutions. Regarding interactivity, the model features camera perspective control and text-driven event generation. According to author @kwangmooyi, the technical implementation utilizes Depth Anything and point cloud technologies, focusing on four core capabilities: interactivity, consistency, stability, and camera movement.

Why it matters

By combining a massive 15B parameter count with open weights, AlayaWorld provides a robust foundation for long-horizon, playable video world generation. Its advanced features, such as text-driven event insertion and camera control, further expand the potential of video models in virtual environment exploration and dynamic content creation.

2026-07-22 ~ 2026-07-23 · 5 related posts

• AlayaWorld Open-Sources Video World Model Supporting 720p/24fps Streaming — iamaliveix · 2026-07-22
• AlayaWorld open-sources a 720p, 24 FPS video world model with camera control — fruesome · 2026-07-22
• [source] AlayaWorld brings interactive long-horizon video world models to 24-fps generation — AlayaLab · 2026-07-22
• [source] AlayaWorld open-weights video world model hits 15B parameters at 24 FPS — kwangmoo_yi · 2026-07-23
• Alibaba Open-Sources AlayaWorld: 15B Model for Playable Video World Generation at 720p — kwangmoo_yi · 2026-07-23