MongoDB - Build AI That Scales
Mini toolsDreamX World
DreamX-World is an interactive world model from AMAP-ML, an AI team at Alibaba AMAP. It generates diverse, high-fidelity worlds that users can explore, control, and transform through action inputs and event prompts. The model is trained on Unreal Engine data, gameplay footage, and real-world videos, using camera estimation, data filtering, progressive training, reinforcement learning, forcing, and distillation to improve action following, interaction consistency, visual fidelity, and efficient inference. The project includes DreamX-World-5B-Cam for short action-controlled generation and DreamX-World-5B for longer 1-minute video generation, with planned 14B and audio-video joint generation variants.
New
Multimodal
Gen 3
Overview
DreamX-World is AMAP-ML’s general-purpose interactive world model for generating controllable, high-fidelity worlds with camera actions and event prompts.
About AutoNavi (Amap / 高德地图)
AutoNavi (Amap / 高德地图) is China's leading digital map, navigation, and location-based services platform, serving over 800 million monthly active users. A fully-owned subsidiary of Alibaba Group since 2014, it develops the Amap mobile app and powers mapping infrastructure for Apple Maps in China, connected vehicles, and enterprise services. Its AMAP-ML research team publishes cutting-edge AI work including world models and geospatial AI.
Industry:
Technology, Information and Internet
Company Size:
1200
Location:
Beijing, Chaoyang District, CN
Website:
amap.com
Tools using DreamX World
No tools found for this model yet.
Last updated: June 16, 2026
