GR-RL is a reinforcement learning framework designed for long-horizon, dexterous manipulation, enabling robots to reliably execute multi-step, high-precision tasks in real-world settings. It is the first in the industry to achieve continuous end-to-end shoe-lacing on a full shoe. The system can automatically retry, adjust the scene, detect misalignment and correct errors, and it generalizes across shoes of different colors, sizes, and materials.

GR-3 is a large-scale vision-language-action (VLA) model. It showcases capabilities in generalization to novel objects, environments, and instructions involving abstract concepts. Furthermore, it can be efficiently fine-tuned with minimal human trajectory data, enabling rapid and cost-effective adaptation to new settings. GR-3 also excels in handling long-horizon and dexterous tasks, including those requiring bi-manual mobile manipulation skills, showcasing its robust and reliable performance.

ByteDexter is a dexterous robotic hand with 21 active degrees of freedom (DoF), fitted with highly sensitive tactile sensors. When paired with its dedicated teleoperation system, it can reproduce complex human hand movements in real time with high fidelity, while enabling seamless hand-arm coordination.