GigaAI

A Chinese AI startup that uses world models to manufacture training data for embodied foundation models.

GigaAI (also known as GigaAI Vision) is a Chinese AI company reportedly founded by Guan Huang. It started out building world models for autonomous driving, the DriveDreamer series, before pivoting to embodied AI. Its core approach is to use world models — models that can generate future video frames — to mass-produce training data, then use that data to train vision-language-action (VLA) models: GigaWorld-0 generates robot manipulation videos and 3D scenes, and the GigaBrain-0 series are the resulting VLA models, with code open-sourced under the open-gigaai organization on GitHub. The company has also released Maker H01, a robot with dual arms on a mobile base. Its website states it closed a RMB 1 billion Pre-B round in March 2026. It is one of the representative companies pursuing the “world model as data engine” approach.

ExampleGigaBrain-0 is trained on real-robot data supplemented with video generated by GigaWorld-0; the company's website says it topped the RoboChallenge leaderboard in February 2026.

Also called

GigaAI Vision

Related

GigaBrain-0 · GigaWorld-0 · World Model · Vision-Language-Action Model · Synthetic Data · RoboChallenge

Sources

极佳科技官网 (Chinese)
GigaAI GitHub organization (open-gigaai)
GigaBrain-0 (arXiv 2510.19430)

As of

2026-03