Reachy Mini

Home > Hardware > Misc > Reachy Mini

Overview

Reachy Mini is an open-source desktop robot co-developed by Hugging Face and Pollen Robotics. Starting at $299, it provides an AI robotics experimentation platform accessible to everyone.

Item	Content
Development	Hugging Face + Pollen Robotics
License	Apache 2.0
Type	Desktop companion robot
Height	28cm (11 inches)
Weight	1.5kg (3.3lbs)
DoF	6 (head) + 360-degree body rotation
Price	$299 (Lite) / $449 (Wireless)
Release	Summer 2025 (Lite), Fall 2025~2026 (Wireless)

Key Significance

Democratization of AI Robotics

Reachy Mini’s most important contribution is opening up the combination of AI and physical robots to general users.

Low-cost Accessibility: At $299, accessible to students, hobbyist developers, and educators alike
Fully Open-Source: Hardware design, software SDK, simulation environment all public
Hugging Face Ecosystem Integration: Immediate access to 1.7M+ AI models and 400K+ datasets
DIY Learning Experience: Hands-on STEM education through 2-3 hour assembly process
Embodied AI Platform: Implementing voice, vision, language models through physical robot

At CES 2026, NVIDIA CEO Jensen Huang demonstrated connecting DGX Spark and Reachy Mini as a “personal office R2D2”. This showed that locally-running AI agents can interact in real-time through physical robots. [1]

Pollen Robotics Background

Company History

Item	Content
Founded	2016, Bordeaux, France
Founders	Matthieu Lapeyre, Alice Coucke, Pierre Rouanet (former Inria researchers)
Mission	Creating open, ethical, and accessible robots
Funding	EUR 2.5M (~$2.8M)
Acquisition	April 14, 2025, acquired by Hugging Face [2]

Key Product History

Poppy (early): One of the first 3D-printed open-source humanoid robots
Reachy 1 (2020): ANA XPrize 2nd place
Reachy 2 (current): $70,000 humanoid, used by research institutions including Cornell/CMU
Reachy Mini (2025): Desktop version for democratization

With Hugging Face’s acquisition, approximately 20 Pollen Robotics employees joined, making this Hugging Face’s fifth acquisition (after Gradio, XetHub, etc.) and first entry into hardware. [2]

Lite vs Wireless Version Detailed Comparison

Item	Lite ($299)	Wireless ($449)
Compute	External (Mac/Linux required)	Built-in Raspberry Pi 5
Connection	USB wired only	WiFi + wired (Ethernet via USB-C adapter)
Power	Wired power	Battery + wired
Microphones	4	4
Speaker	5W	5W
Camera	Wide-angle camera	Wide-angle camera
Accelerometer	No	Yes
Autonomy	PC dependent	Independent operation possible
Shipping Start	Summer 2025	Fall 2025~2026

Version Selection Guide

Lite Recommended: Developers/educators who already have a powerful PC and are okay with wired connection
Wireless Recommended: Users needing standalone demos, mobility, or familiar with Raspberry Pi projects

Hardware Specs

Physical Specifications

Item	Spec
Height	28cm (11 inches)
Width	16cm (6.3 inches)
Weight	1.5kg (3.3lbs)
Head DoF	6 (pan, tilt, roll, etc.)
Body Rotation	360-degree continuous rotation
Antennas	2 (for animation expression)

Sensors and Output

Component	Spec
Eyes	Expressive LED animations
Camera	Wide-angle camera (for vision processing)
Microphones	4-channel array
Speaker	5W output
Accelerometer	Wireless version only

Assembly

Reachy Mini is sold as a DIY kit, typically taking 2-3 hours to assemble. The assembly process itself provides educational value.

Software Ecosystem

Python SDK

from reachy_mini import ReachyMini

# Auto connect (USB/localhost or network auto-detect)
reachy = ReachyMini()

# Or force connection mode
reachy = ReachyMini(connection_mode="localhost_only")
reachy = ReachyMini(connection_mode="network")

# Head movement control
reachy.head.look_at(x=0.5, y=0, z=0.3)

# Eye animation
reachy.eyes.set_emotion("happy")

# Camera access
frame = reachy.camera.get_frame()

Supported Programming Languages

Language	Status
Python	Currently supported (full SDK)
JavaScript	In development (coming soon)
Scratch	In development (coming soon)

Hugging Face Hub Integration

Reachy Mini is deeply integrated with the Hugging Face ecosystem:

1.7M+ AI models immediately accessible
400K+ datasets available
Hugging Face Spaces for app sharing and community collaboration
SmolVLM2-2.2B-Instruct default local vision model

# Local vision processing with SmolVLM2
# --local-vision flag processes frames periodically

App Development and Deployment

# App creation tool
reachy-mini-app-assistant create

# Deploy app to Hugging Face Space to share with community

Educational Applications

STEM Education Application Areas

Field	Content
Programming	Python basics, AI integration, asynchronous programming
AI/ML	Computer vision, speech recognition, LLM integration
Robotics	Motion planning, sensor fusion, control systems
HRI	Human-robot interaction, emotion expression, social robots

Target Ages and Environments

Environment	Recommendations
High School+	Independent Python-based learning possible
Middle School	Use with teacher or mentor guidance
Elementary	Suitable after Scratch support (in development)
University/Research Labs	AI robotics, motion planning, autonomous decision-making experiments

Educational Advantages

Hands-on Learning: Experience abstract AI concepts through physical robot
Visual Feedback: Immediate robot state confirmation via LED eyes and antennas
Progressive Difficulty: Web dashboard → Python SDK → Advanced AI integration
Community Learning: Reference and share other projects on Hugging Face Spaces

CES 2026 Demonstration

NVIDIA DGX Spark + Reachy Mini

At CES 2026, Jensen Huang demonstrated connecting Reachy Mini with NVIDIA DGX Spark. [1]

Item	Content
AI Brain	NVIDIA DGX Spark (100B parameter model local execution)
Physical Avatar	Reachy Mini
Model Routing	Real-time intent-based router (privacy → local, complex reasoning → cloud)
Voice	Eleven Labs API integration

Demonstrated Features

Grocery list management (eggs, milk, butter)
Script update delivery via code execution
Hand-drawn sketch to architectural rendering conversion
Real-time home monitoring

“What’s amazing is that this is now completely easy. But just a few years ago, this was impossible, unimaginable.” - Jensen Huang [1]

Purchase and Shipping

Status (as of January 2026)

Shipped: 3,000 units shipped through Hugging Face, Pollen Robotics, Seeed Studio collaboration [3]
Development Period: About 5 months from open-source idea to actual product [3]

Where to Buy

Store	Notes
Pollen Robotics Official	pollen-robotics.com/reachy-mini
Seeed Studio	Kits and parts
Hugging Face Hub	Software and models

System Requirements

Lite Version

Item	Requirement
OS	macOS 10.15+ or Linux (Ubuntu 20.04+ recommended)
Python	3.8+
Connection	USB
Windows	Coming soon

Wireless Version

Item	Requirement
Built-in Compute	Raspberry Pi 5
Remote Access	Web browser (any device)
Network	WiFi or USB-C Ethernet adapter

References

Citation Sources

[1] NVIDIA brings agents to life with DGX Spark and Reachy Mini - Hugging Face Blog, January 2026

[2] Hugging Face buys a humanoid robotics startup - TechCrunch, April 14, 2025

[3] Reachy Mini: An Open Journey Built Together - Seeed Studio Blog, January 6, 2026

Reachy Mini

Reachy Mini

Overview

Key Significance

Democratization of AI Robotics

Pollen Robotics Background

Company History

Key Product History

Lite vs Wireless Version Detailed Comparison

Version Selection Guide

Hardware Specs

Physical Specifications

Sensors and Output

Assembly

Software Ecosystem

Python SDK

Supported Programming Languages

Hugging Face Hub Integration

App Development and Deployment

Educational Applications

STEM Education Application Areas

Target Ages and Environments

Educational Advantages

CES 2026 Demonstration

NVIDIA DGX Spark + Reachy Mini

Demonstrated Features

Purchase and Shipping

Status (as of January 2026)

Where to Buy

System Requirements

Lite Version

Wireless Version

References

Citation Sources

Official Documentation

Development Resources

See Also