01
RESERVE NOW
Lock in Founder Price.
100% refundable before dispatch.

Natural conversation, autonomous following, and companionship for the whole family.
Reserve your spot to co-create the open RL platform.
01
Lock in Founder Price.
100% refundable before dispatch.
02
Priority shipping.
Discord Co-Creator access — shape the SDK with us
03
Ships September 2026.
You'll get early beta access.
LIMITED Founder Edition
M1 DUAL-MODE HUMANOID · OPEN PLATFORM · ON-DEVICE AI
BUY NOWM1
ENGINEERED FOR REAL-WORLD RL DISCOVERY.
ENGINEERED FOR REAL-WORLD RL DISCOVERY.
01Train locomotion and balance policies with reference workflows in Isaac Gym. Designed for PyTorch/ONNX model export and sim-to-real pipelines via the Reinforced Imitation Framework.
02Gain direct access at every layer with the Python SDK or the M1 Spine SDK for low-latency control. Designed for ROS 2 integration and Linux + OTA workflows.
03Equipped with dual binocular cameras for high-precision local depth perception. Supports on-device SLAM, obstacle negotiation, and visual spatial reasoning with no cloud latency required.
04Execute neural network inference locally on the 10 TOPS NPU (BPU). Stream motor torque, thermal states, 6-axis IMU data, and bus metrics into your training logs or ROS 2 environment.
05Actuators engineered for dynamic balance, low inertia, and high impact recovery. Designed to withstand real-world trial-and-error policy testing.
06A compact 50 cm, 2.7 kg (≈ 6 lb) desktop humanoid. Transitions between bipedal and wheeled modes to support diverse control policies in tight spaces.
SWIPE TO EXPLORE 01 — 06
05 / TECHNICAL SPECIFICATIONS
06 / FAQ
The essentials, answered clearly.
Yes — full refund any time before dispatch.
Yes — US, EU, UK, Canada, Australia, Japan, and Singapore at launch. Import duties may apply.
1-year standard warranty, extendable to 3 years. VIPs receive 12-hour priority response.
Yes. M1 provides reference workflows designed specifically for Isaac Gym. Policies trained in PyTorch can be easily exported via ONNX and deployed to the physical robot through the Reinforced Imitation Framework for efficient sim-to-real transfer. Example training scripts and reference policies are available directly through the private developer Discord community.
Yes. M1 features an integrated 10 TOPS NPU (BPU) optimized for local neural network inference, processing real-time spatial data directly from the dual binocular cameras. Developers can deploy custom models via the standard ONNX toolchain, and the Joyin engineering team is available to assist with model quantization recipes and operator optimization.
M1 runs a flexible Linux host environment with native ROS 2 message-bus support, allowing developers to connect custom software components and external nodes. Detailed I/O specifications, hardware expansion capabilities, and pinout diagrams will be fully detailed in the forthcoming hardware developer guide.
To safeguard the robot's high-torque actuators during dynamic policy testing, M1 features built-in hardware force limits by default. This ensures safe operation and long-term durability during aggressive movement, push recovery, and contact-rich reinforcement learning experiments.