GigaBrain-0.5M*
$ 0
GigaBrain-0.5M* is GigaAI’s robot action model that learns from a world model’s predictions. It reliably finishes long chores like folding laundry and making espresso.
Available on backorder
Brain Score
1
Specifications and details:
| Nationality | China |
|---|---|
| Website | https://gigaai.cc/ |
| Model type | Foundation Model |
| Manufacturer | GigaAI |
| Release date | February 12, 2026 |
Description
GigaBrain-0.5M* is a vision-language-action model from GigaAI, a Chinese embodied AI company. It builds on GigaBrain-0.5, which learned from more than 10,000 hours of robot manipulation data. GigaAI also used its own GigaWorld world model to create much of that training data. The new model goes further, because it learns through reinforcement learning guided by world model predictions. The company calls this method RAMP.
With RAMP, the robot policy sees what the world model expects to happen next and how valuable that future looks. Then the team collects real robot runs, adds human corrections and refines both models together. As a result, GigaAI reports near-perfect success on long tasks such as folding laundry, packing boxes and making espresso. The company shared the technical report and project page with demo videos. However, it has not posted GigaBrain-0.5M* weights, although earlier and later GigaBrain models are open. The model sits between GigaBrain-0.1 and GigaBrain-0.7 in GigaAI’s fast release schedule.
New! 2026 Humanoid
Robot Market Report
198 pages of exclusive insight from global robotics experts — uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
now Google DeepMind
Contact Humanoid.guide
Website: https://gigaai.cc/blog/gigabrain07
