AI Tool

Microduck Review

Microduck is a 25 cm open-source bipedal AI robot designed for physical AI experimentation, reinforcement learning, and play.

shipped Aug 28, 2026paid

Domain rating56Monthly visits2.3K/mo

Why it matters

125 cm tall, 800-gram open-source bipedal robot with 15 motors.

2Pre-orders opened August 27, 2026, with an introductory price of $399.

3Features an open-source software stack for robot control, simulation, and reinforcement learning.

4Equipped with a camera, LiDAR, two IMUs, and an articulated beak for grasping.

About Microduck

Business Model

Open Source

Target Audience

AI builders and makers

Pricing Plans

Microduck

$399 / one-time

• Open source
• Pre-order
• Ships before Christmas 2026

Charger pack

$39 / one-time

• Dual charger
• 2x batteries

Dev pack

$119 / one-time

• 3x spare motors
• 5x motor cables
• 2x batteries
• dual charger

Accessory pack

$39 / one-time

• Laser pointer
• NFC polaroid
• 2x rollers
• ball

GitHubOpen Source

overview

What is Microduck?

Microduck is a physical AI robot tool developed by Pollen Robotics (acquired by Hugging Face in April 2025) that enables developers, AI builders, and makers to experiment with physical AI and reinforcement learning. It is a 25 cm (9.8 inches) tall, 800-gram (1.8 pounds) open-source bipedal robot equipped with 15 motors, a camera, LiDAR, two motion sensors (IMUs), and an articulated beak for grasping objects. It comes with seven pre-built capabilities, including walking, sitting, crouching, getting up from falls, kicking a ball, and picking up objects, with an optional accessory pack for roller skating. Its primary use cases include physical AI and reinforcement learning, education and accessibility in robotics, and play and exploration.

features

Key Features of Microduck

Microduck integrates a comprehensive set of hardware and software features designed for physical AI experimentation and reinforcement learning. Its open-source nature allows for extensive customization and community contributions.

25 cm tall, 800-gram open-source bipedal robot.
15 motors for whole-body movement and articulated beak for grasping.
Integrated wide-angle front camera, 8x8 ToF LiDAR, and two built-in IMUs.
Microphone and speaker for audio interaction, and two NFC antennas.
Powered by Rockchip RK3566 processor with AI accelerator, 1GB RAM, 32GB storage.
Supports Wi-Fi and Bluetooth connectivity.
Open-source software stack for robot control, simulation, and reinforcement learning (e.g., PPO, ONNX models).
Sim-to-real deployment capabilities for training behaviors in simulation and deploying them on the physical robot.
Self-recovery mechanism to get up from falls, facilitating iterative learning.
Removable 2,600mAh battery providing approximately 1 hour of operation.

use cases

Who Should Use Microduck?

Microduck is designed for individuals and groups interested in hands-on AI and robotics development, offering a platform for both learning and advanced experimentation.

Developers: For building and deploying new AI behaviors using reinforcement learning and the open-source SDK.
AI Builders: For experimenting with physical AI, training new behaviors in simulation, and deploying them in the real world.
Makers: For engaging with robotics and AI in an accessible, pre-built format, suitable for tinkering and customization.
Educators and Students: For democratizing access to robotics and AI, enabling direct experimentation without extensive prior coding knowledge.
Researchers: For exploring advanced topics in bipedal locomotion, object manipulation, and sim-to-real transfer learning.

how to use

How to Use Microduck

Microduck is designed for immediate use out of the box, with options for advanced training and customization through its open-source ecosystem. Users can control it with a gamepad or develop custom AI policies.

1Unbox and Power On: Activate the Microduck, which generates a unique audio identity upon first activation.
2Initial Play: Control the robot using the included gamepad to explore its pre-built capabilities like walking, sitting, and kicking.
3Simulation Training: Utilize the open-source SDK and simulation environment (available on GitHub) to train new behaviors using reinforcement learning (e.g., PPO, ONNX models).
4Deploy to Robot: Deploy the trained AI policies from the simulation directly onto the physical Microduck robot.
5Refine and Publish: Fine-tune behaviors on the physical robot and contribute new policies to the community via GitHub.
6Access Community Support: Engage with other users and developers through the dedicated Discord channel for assistance and collaboration.

pricing

Microduck Pricing & Plans

Microduck is available for pre-order with an introductory price of $399 before taxes and shipping. Optional accessories are sold separately to enhance functionality and development capabilities. Bulk pricing and dedicated support are offered for orders of 10 or more units.

Microduck: $399 (one-time purchase for the robot in Cream, Graphite, Lavender, or Sky colorways).
Charger pack: $39 (one-time purchase for additional charging capabilities).
Dev pack: $119 (one-time purchase for development-focused accessories).
Accessory pack: $39 (one-time purchase, includes rollers for skating, a ball, a laser pointer, and NFC tags).

Pros

+Open-source software stack for full control and customization, including SDK and RL tools on GitHub.
+Compact 25 cm size and self-recovery from falls make it safe and manageable for experimentation.
+Supports sim-to-real deployment, allowing behaviors trained in simulation to be deployed on the physical robot.
+Accessible entry point into physical AI and reinforcement learning for developers, AI builders, and makers.
+Equipped with 15 motors, camera, LiDAR, and IMUs for diverse sensory and motor capabilities.
+Introductory price of $399 makes it a budget-friendly option compared to other robotics platforms.

Cons

−Battery life is approximately 1 hour, requiring frequent recharging for extended use.
−Initial user reviews noted some lagginess in the simulator, which may impact development workflow.
−Requires an understanding of reinforcement learning concepts for advanced behavior training.
−Optional accessories (e.g., Dev pack, Accessory pack) are sold separately, increasing the total cost for full functionality.
−As a new product, long-term community support and extensive third-party integrations are still developing.

Similar Tools

Microduck vs Competitors

Microduck is positioned as a budget-friendly, open-source bipedal robot, differentiating itself from more complex or DIY-intensive alternatives by offering an out-of-the-box experience with robust sim-to-real capabilities.

Upkie↗

Upkie is a wheeled biped robot that balances on two wheels while using articulated legs to traverse uneven terrain, offering Gymnasium-compatible reinforcement learning environments.

While Microduck is a pure biped, Upkie is a wheeled biped, which means its locomotion dynamics are different. However, it provides a complete open-source platform for reinforcement learning with detailed build documentation, requiring more DIY effort than Microduck's out-of-the-box experience.

MEVITA↗

MEVITA is an open-source bipedal robot designed for minimal viable configuration, built entirely from readily available e-commerce components using sheet metal welding.

MEVITA offers a truly open-source hardware and software stack for bipedal reinforcement learning, similar to Microduck's philosophy. The trade-off is that MEVITA requires significant DIY assembly and fabrication (sheet metal welding), whereas Microduck is a pre-built product.

See every Microduck alternative, compared→