Hy-Embodied Open Source Model
Focused on developing open-source embodied foundation models to advance robots toward human-like, general-purpose embodied intelligence capable of complex, efficient, and flexible interaction with the physical world.
Hy-Embodied-RxBrain-1.0

Built on Hy-Embodied-VLM, the embodied world model jointly models world understanding, reasoning, planning, and action prediction. Trained on 50K+ hours of high-quality embodied data, it unifies text, image, and video understanding and generation, providing high-level intelligence for embodied decision-making and action generation.

Check the details

Built on Hy-Embodied-VLM, the embodied world model jointly models world understanding, reasoning, planning, and action prediction. Trained on 50K+ hours of high-quality embodied data, it unifies text, image, and video understanding and generation, providing high-level intelligence for embodied decision-making and action generation.

Check the details
Hy-Embodied-0.5-VLA

An end-to-end embodied AI full-stack system covering the entire pipeline from proprietary data collection and large-scale human demonstration datasets to UMI-based pre-training, supervised fine-tuning, offline reinforcement learning, and real-world deployment.

Check the details

An end-to-end embodied AI full-stack system covering the entire pipeline from proprietary data collection and large-scale human demonstration datasets to UMI-based pre-training, supervised fine-tuning, offline reinforcement learning, and real-world deployment.

Check the details
Hy-Embodied-0.5-X

Bridging the gap from "perception" to "action", the model leverages proprietary first-person robot and robotic arm operation data to optimize core embodied AI capabilities, including manipulation understanding, spatial reasoning, action prediction, risk assessment, multimodal understanding, and long-horizon planning.

Check the details

Bridging the gap from "perception" to "action", the model leverages proprietary first-person robot and robotic arm operation data to optimize core embodied AI capabilities, including manipulation understanding, spatial reasoning, action prediction, risk assessment, multimodal understanding, and long-horizon planning.

Check the details