Meta builds humanoid robot prototype in 3 months, already developing Max 2 with improved capabilities

Reviewed byNidhi Govil

8 Sources

Share

Meta partnered with Chinese robotics firm Fourier to develop Max, a 48kg humanoid robot prototype, in just three months. The 1.5-meter-tall robot runs on Nvidia's Jetson Thor chip and features five-fingered hands, facial expressions, and voice capabilities. Meta is already working on Max 2 with enhanced torso design, better hands, and stair navigation sensors.

News article

Meta Accelerates Humanoid Robot Development with Fourier Partnership

Meta has developed a humanoid robot prototype called Max in collaboration with Shanghai-based robotics company Fourier in approximately three months

1

2

. The Meta humanoid robot weighs about 48kg (106 pounds) and stands roughly 1.5 meters tall, making it comparable in size to an average adult

5

. This internal research project marks a significant step in Meta's AI robotics ambitions, though the company clarified it has no plans to sell these units and the work remains focused on experimentation

2

.

The Max prototype runs on the Nvidia Jetson Thor chip, providing the processing power needed for complex robotic functions

3

. The humanoid robot prototype features a display that mimics facial expressions, built-in speakers for audio output, and five-fingered dexterous hands designed to manipulate small objects including bottles and toys

1

. Meta's design requirements emphasize natural human interaction, including voice recognition from approximately five meters away and gesture interpretation from around three meters

3

. The robot is also expected to maintain eye contact during interactions and communicate its operating state, indicating whether it is listening, processing, speaking, or uncertain about a task

3

.

Deep Collaboration Between Meta and Fourier in Shanghai

The collaboration with Fourier involved substantial commitment from Meta's robotics leadership. Li-Chen Miller, Meta's vice president of product for robotics, spent almost two months at Fourier's factory in Shanghai working directly with the Chinese robotics firm

2

5

. Meta's robotics team traveled to China specifically for this project, demonstrating the company's serious investment in developing physical AI capabilities

4

. The prototype was showcased to Meta employees at an internal robotics summit in late August

2

.

Max represents a customized platform drawing on hardware and engineering experience from three generations of Fourier robots, rather than simply a modified version of existing models like Fourier's GR-3

3

. Fourier unveiled its GR-3 humanoid in August 2025, marketing the 165cm robot as a care-bot and companion for uses including support for older adults

2

5

. The partnership leverages Fourier's experience with humanoid robots that integrate hearing, vision, and touch capabilities

3

.

Second-Generation Max 2 Already in Development

Meta began work on the second-generation humanoid robot, codenamed Max 2, in September—just months after completing the first prototype

1

4

. The updated version will feature a redesigned torso, improved hands for better object manipulation, a projector, and additional sensors specifically intended to help with stair navigation

1

2

. These enhancements suggest Meta is working towards a robot capable of interacting more naturally with household environments rather than simply demonstrating basic humanoid movements

4

.

The rapid iteration from Max to Max 2 indicates Meta's aggressive development timeline, though the company has not set a launch date for any commercial humanoid product

2

5

. Meta has been recruiting robotics engineers in Shanghai and Shenzhen, with positions covering prototype development and preparations for potential large-scale manufacturing

3

. This hiring activity suggests the company is building infrastructure for more than just research prototypes, even as it maintains that current work is experimental.

Connecting AI Agent Muse to Physical Robotics

Meta views humanoid robots as a physical AI form for Muse, its AI agent, with the goal of enabling these robots to perform household chores

2

5

. According to reports, Meta CTO Andrew Bosworth has discussed humanoid robots internally as a potential physical embodiment of the company's AI agent Muse

3

. This represents a shift from AI that exists purely in digital interfaces to AI that can interact with the physical world and complete tangible tasks

4

.

Nvidia, whose Jetson Thor chip powers Max, agreed this week to invest $3 billion in Meta and expand their work on Muse

2

5

. This substantial investment signals confidence in Meta's robotics direction and strengthens the partnership between the two companies. The collaboration positions NVIDIA's hardware as central to Meta's robotics ambitions, similar to how Agility's Digit 5 humanoid will also run on Nvidia hardware

2

.

Building Robotics Capabilities Through Strategic Acquisitions

Meta has been systematically expanding its robotics team through acquisitions and hiring. In May, the company acquired Assured Robot Intelligence, a San Diego startup that develops AI software for robots

1

5

. The acquisition adds expertise in developing AI systems that allow robots to understand their environments and perform physical tasks

3

. The team from Assured Robot Intelligence joined Meta Superintelligence Labs, which houses the group leading Meta's humanoid work

2

5

.

These moves follow Bloomberg's report from last year that Meta created a new team tasked with building robot hardware with the eventual goal of performing household chores

1

. The company is entering a crowded field that includes Tesla's Optimus robots and Chinese rival Unitree, which is working on AI that lets its robots improve themselves

2

. However, a robot that can pick up a bottle remains far from one that can reliably manage household tasks

4

. Questions remain about the robots' ability to perform complex household chores and operate safely around people, indicating that while Meta has made progress, significant technical challenges lie ahead before these prototypes could become practical home assistants

3

.

© 2026 TheOutpost.AI All rights reserved