KORTHOS / ECOSYSTEM / PRODUCT

Astribot S1

About

About Astribot S1

Astribot S1 is a bimanual humanoid-style robot for embodied-AI research and service-task exploration. Its software-hardware architecture couples onboard AI with two dexterous arms, seven degrees of freedom per arm, a 5 kg horizontal payload and an approximately 170 cm form factor.

The platform provides APIs, development guidance, visual programming, simulation support and VR teleoperation for researchers building manipulation policies. S1 can fold, pour and handle everyday objects in demonstrations, serving as the physical robot and data-collection interface for learning, control and interaction experiments.

03 JUN 2026

Astribot is the latest Chinese unicorn

Capabilities

Source-backed capabilities

  • Capability areas: Mobile manipulation, Force control
  • Workflows: Robot learning data
  • Operating environments: Research lab
  • Target markets: Robotics researchers, AI labs and embodied AI teams.
  • Deployment model: Research platform available through Astribot's product channels.
  • Object manipulation

    Research demonstration - 2025-12-09 - Mind to Hand evaluates embodied reasoning that connects task intent to manipulation actions on Astribot S1.

    Source
  • Bimanual manipulation

    Research demonstration - 2025-12-09 - Training combines vision-language data with cross-embodiment robot data and trajectories collected on the Astribot S1 bimanual mobile manipulator.

    Source

Research

  • Astribot releases SmoothRL for online learning on S1

    2026-09-04 - Research Publication - SmoothRL separates committed, executed and discarded actions so policy gradients use only actions that physically ran. After 250 episodes on S1, reported success reached 94% for dynamic tossing, 83% for pen capping and 90% for box opening. The training loop supports VR takeover and residual joystick correction while the robot continues operating.

    Source
  • RynnBrain 1.1: Towards More Capable and Generalizable Embodied Foundation Model

    2026-07-20 - Research Publication - The authors present RynnBrain 1.1, a family of embodied foundation models spanning 2B, 9B, and 122B-A10B scales. The authors also develop RynnBrain-VLA with a unified cross-embodiment action space and embodiment-specific masking, and deploy it on Unitree G1, Astribot-S1, and Tianji-Wuji. Real-robot experiments show that RynnBrain-initialized policies outperform Qwen-based and representative generalist VLAs, while joint multi-task and multi-embodiment training improves process scores and success rates over per-task training.

    Source
  • A Glimpse into Long-term Physical Coexistence with Intelligent Robots

    2026-07-13 - Research Publication - The authors introduce PHILIA, a multi-robot agent built around a robot gateway abstraction. The authors validate the architecture on Astribot S1 robots while designing the robot gateway contract to support future heterogeneous robot platforms through a shared capability interface for observation, task execution, navigation, speech playback, status monitoring, and task cancellation.

    Source
  • Mind to Hand: Purposeful Robotic Control via Embodied Reasoning

    2025-12-09 - Research Publication - The authors introduce Lumo-1, a generalist vision-language-action model that connects robot reasoning with physical action. Training combines vision-language data with cross-embodiment robot data and trajectories collected on the Astribot S1 bimanual mobile manipulator. Experiments report improved embodied vision-language reasoning for generalist robotic control.

    Source

Timeline

  • Astribot releases SmoothRL for online learning on S1

    2026-09-04 - Research Publication - SmoothRL separates committed, executed and discarded actions so policy gradients use only actions that physically ran. After 250 episodes on S1, reported success reached 94% for dynamic tossing, 83% for pen capping and 90% for box opening. The training loop supports VR takeover and residual joystick correction while the robot continues operating.

    Source
  • Astribot unveils Astribot S1 as its next-generation AI robot assistant

    2024-04-24 - Product Launch - Astribot publicly introduced Astribot S1 as its next-generation AI robot assistant with a focus on speed, dexterity, and precision. The launch established Astribot's company-defining humanoid platform and moved the firm from stealth into public embodied-robot positioning. This is the clearest early product milestone for the company before its later embodied-reasoning model work.

    Source

Context

Market context

Astribot S1 is a wheeled bimanual mobile manipulation platform focused on household and service-style embodied AI tasks.

  • Target environment: Research labs, development workshops and controlled service environments.
  • Workflow context: Bimanual manipulation, object handling and VLA model training for real-world embodied AI tasks.
  • Customer context: Robotics researchers, AI labs and embodied AI teams.
  • Deployment model: Research platform available through Astribot's product channels.
  • Commercial maturity: Prototype-stage research platform with a strong public demonstration track record.
  • Adoption constraints: Still pre-commercial for broad workplace deployment. The main role is as a research and model-training platform.
  • Market position: One of the more visible wheeled bimanual platforms in the embodied AI research space, alongside AgiBot G2 in the mobile manipulation category.
  • Adjacent products: AgiBot G2, DexLeap Mobile Manipulator

Tags

  • Peer Group: Wheeled dual-arm humanoid mobile manipulator, Mobile manipulation robot
  • Workflow: Robot learning data
  • Capability: Mobile manipulation, Force control
  • Deployment Environment: Research lab
  • Stack Layer: Finished robot platform

Comparable Systems