A California charter school paid $500,000 for two humanoid robots and installed them in a strip mall classroom. Students were asked to interview Nikola Tesla. The robot interrupted them mid-sentence, spoke too fast, and had to repeat its introduction three times. The number one word students used to describe it: "creepy."
The robot is called Ameca. It is made by a British company called Engineered Arts. And despite the school billing it as the "world's most advanced AI-powered humanoid robot," Ameca cannot walk. It does not pick up objects. It does not navigate rooms. It stands on a weighted base and talks.
That gap between the marketing and the machine is worth understanding — because Ameca keeps showing up in schools, museums, and headlines, and almost nobody explains what it actually is.
What Ameca Actually Is
Ameca is a social humanoid robot. That category is important. Social humanoids are built for human-robot interaction — conversation, facial expression, engagement. They are not built for tasks. They do not carry boxes, assemble parts, or walk across factory floors.
Engineered Arts, a company headquartered in Falmouth, Cornwall, UK, first revealed Ameca in December 2021. Its public debut was at CES 2022 in Las Vegas, where it became widely shared online for the uncanny realism of its facial movements. The company describes it as a platform for AI research and human-robot interaction development — not a general-purpose worker.
Ameca's standout feature is expressiveness. Its face uses Engineered Arts' proprietary Mesmer technology — dozens of individually actuated motors that replicate the specific muscle movements behind human expressions. The result is a robot that can raise one eyebrow skeptically, compress its lips, blink with slight delay, and yawn. These are not programmed animations. They are generated in real time based on conversational context.
The body is deliberately designed to read as non-human. Grey rubber skin on the face and hands. Exposed mechanical joints at the elbows and shoulders. A transparent skull dome with interior LED lighting. The intent is to avoid the uncanny valley effect — that specific discomfort that emerges when a robot looks almost human but not quite. Ameca leans into the "clearly a machine" aesthetic while maximizing facial realism.
Price: $100,000 to $500,000 — What Each Configuration Buys
Engineered Arts does not publish prices on its website. Based on multiple industry sources verified as of early 2026, here is how the pricing breaks down by configuration:
| Configuration | Estimated Price | Use Case |
|---|---|---|
| Head only | $25,000 – $100,000 | Reception desk, kiosk |
| Half body | ~$100,000 | Exhibition, event demos |
| Full body (standard) | ~$130,000 – $250,000 | Research, museum install |
| Full body (custom enterprise) | Up to $500,000 | Custom deployment, NDA projects |
Additional costs are not optional — professional installation by Engineered Arts engineers is required, plus ongoing software licensing and maintenance. The purchase price is not the total cost.
For context: a Unitree G1, which actually walks, runs, and is being evaluated for factory work, starts at $16,000. That comparison is not quite apples to apples — Ameca and Unitree G1 are solving different problems — but the 15x price gap is worth noting when evaluating what $250,000 buys in the humanoid market today. For a fuller picture of what working humanoids cost in 2026, see 6 Humanoid Robots That Are Real, Priced, and Shipping in 2026.
Full Specs: 61 Degrees of Freedom, Tritium Software, ChatGPT
Here are Ameca's verified technical specifications as of 2026:
| Spec | Value |
|---|---|
| Height | 180 cm (6 ft 2 in) |
| Weight | ~49–60 kg |
| Degrees of Freedom (total) | 61 actuated DoF |
| Facial expression DoF | 27 dedicated to face alone |
| Facial expressions | 60+ distinct expressions pre-loaded |
| Cameras | Binocular eye-mounted + chest camera |
| Audio | Binaural microphones + studio-quality chest mic |
| Languages | 50+ |
| AI integration | GPT models (ChatGPT), third-party NLP |
| Software | Tritium (custom Linux, C++/Rust/Python) |
| Connectivity | Cloud-connected, remote control supported |
| Locomotion | None — stationary platform on weighted base |
| Operating environment | Indoor only, 10°C – 30°C |
The Tritium software suite is where Ameca's flexibility lives. It runs on a custom Linux distribution and allows operators to define custom "Roles" — distinct personas with their own names, languages, conversational styles, and knowledge bases. No coding required for persona creation. The Altus School configured four of these roles: Sage (teacher), Remi (wellness coach), Ari (career planner), and Lexi (translator).
Tritium also supports integration with multiple LLMs — GPT models, Claude, Gemini — and handles speech recognition, text-to-speech, and voice synthesis as a unified pipeline. Ameca Generation 3, released at ICRA 2025, updated these integrations and added subtler micro-expression control and faster response latency.
Where Ameca Is Actually Deployed
As of early 2026, confirmed Ameca installations include:
- National Robotarium, Edinburgh, UK — permanent installation since 2024, public engagement and research
- Museum of the Future, Dubai — part of the museum's robotic family, visitor interaction
- Computer History Museum, Mountain View, California — AI exhibit
- Heinz Nixdorf MuseumsForum, Paderborn, Germany — ChatGPT integration for visitor conversations
- Copernicus Science Center, Warsaw, Poland — science education
- European Space Agency — ambassador role at the 18th European Space Conference, January 2026
- Altus Charter School, San Diego, California — teaching pilot program, installed January 2026
The pattern is consistent: museums, science centers, research labs, and flagship events. The Altus School deployment is among the first attempts to use Ameca as a daily classroom tool rather than an exhibit or demo.
Social Humanoid vs. Working Humanoid — The Distinction That Matters
The humanoid robot market in 2026 has two very different categories, and they are frequently confused in coverage.
Working humanoids — Unitree G1, Figure 03, Boston Dynamics Atlas, Agility Digit — are built for physical tasks. They walk, carry loads, navigate stairs, pick objects off shelves. Their design priority is locomotion, manipulation, and real-world durability. The G1 starts at $16,000. Figure 03 is heading toward factory floors at BMW. These machines are measured by task completion rates, payload capacity, and hours of uptime. For a breakdown of this category, see Unitree's Humanoid Robot Lineup Explained: $4,900 to $650,000.
Social humanoids — Ameca, Hanson Robotics' Sophia, SoftBank's Pepper — are built for interaction. Their design priority is facial expressiveness, conversational fluency, and approachability. They are measured by engagement quality, dwell time, and how comfortable humans feel talking to them. Ameca does not compete with Unitree on tasks. It competes with nothing on facial realism — that is genuinely its category-defining strength.
The confusion happens when a social humanoid is evaluated against working-humanoid expectations, or vice versa. Calling Ameca "the world's most advanced humanoid robot" — as both Engineered Arts and the Altus School did — collapses this distinction. Advanced at what, exactly, matters a great deal at $250,000 per unit.
My Take
Ameca is genuinely impressive for what it is. The facial engineering is not hype — 27 degrees of freedom dedicated to a single face, generating micro-expressions in real time based on conversational context, is a real technical achievement. If you need the most expressive social robot on the market for a museum, research lab, or brand activation, nothing else is close.
The Altus School purchase is a different question. Three independent university researchers — from UCL, Monash, and MIT — called it ineffective, premature, and potentially harmful to children. The school's own early data showed the robot's first lesson went poorly, students found it unsettling, and there is no set timeline or benchmark for the pilot. Half a million dollars is a significant commitment for a system with no confirmed educational outcome evidence.
The category confusion is the real problem here. Ameca is being sold into schools as a teaching partner. It is, by its own manufacturer's description, a platform for human-robot interaction research. Those are not the same thing. One is a product ready for daily classroom use. The other is infrastructure for studying what robots might eventually become.
- Ameca is a social humanoid robot — built for conversation and expression, not physical tasks. It cannot walk.
- Price ranges from $100,000 (head only) to $500,000 (full custom enterprise). The Altus School paid approximately $250,000 per unit.
- 61 total degrees of freedom, with 27 dedicated to facial expressions alone — the highest facial DoF of any commercially available robot.
- Runs on Tritium software with ChatGPT/GPT model integration and supports 50+ languages.
- Deployed primarily in museums, research labs, and science centers. School use is new and, as of June 2026, without confirmed outcome data.
- Social humanoids and working humanoids are distinct categories. Comparing Ameca to Unitree or Figure is a category error.
FAQ
Can Ameca walk?
No. Ameca is a stationary platform mounted on a weighted base. It has articulated arms, hands, and an expressive upper body, but no functional bipedal locomotion. Engineered Arts has indicated locomotion is on its development roadmap, but no confirmed timeline exists as of 2026.
How much does Ameca cost in 2026?
Engineered Arts does not list prices publicly. Based on verified industry sources as of early 2026, the head-only configuration starts around $25,000–$100,000, a standard full body unit runs approximately $130,000–$250,000, and a fully customized enterprise deployment can reach $500,000. Professional installation and software licensing are additional costs.
What AI does Ameca use?
Ameca runs on Engineered Arts' Tritium software platform, which integrates with third-party AI services including OpenAI's GPT models (ChatGPT), as well as other LLMs like Claude and Gemini. Speech recognition, natural language processing, and text-to-speech are handled as a unified cloud-connected pipeline. The Altus School deployment uses ChatGPT integration specifically.
Is Ameca the same as a working humanoid robot like Unitree or Figure?
No — they belong to different categories entirely. Ameca is a social humanoid optimized for conversation, facial expression, and human engagement. Working humanoids like Unitree G1 or Figure 03 are built for physical tasks: walking, carrying loads, operating in factories. Ameca is best suited for museums, research, and interactive installations. Working humanoids are being deployed in logistics and manufacturing.
Where is Ameca being used in education?
As of 2026, Altus Charter School in San Diego, California is running a pilot with 2 Ameca units installed in its Tierrasanta Resource Center. Richard Bland College in Virginia also has 2 Ameca units used for STEM education and research. Engineered Arts has separately partnered with at least one Southern California school district for supplemental tutoring experiments, though that deployment has not been independently confirmed.
Conclusion
Ameca is the best social humanoid robot available today at a price point that reflects that position. Its facial engineering is category-defining. Its software platform is genuinely flexible. Its real-world track record in museums and research institutions is solid.
Whether it belongs in a K-12 classroom at $250,000 per unit — with no outcome data, no established pedagogy, and researchers from three major universities expressing serious concern — is a different question. The honest answer, as of June 2026, is that nobody knows yet. The Altus pilot is still running, with no defined endpoint.
What is clear is that "humanoid robot" in 2026 covers a much wider range of machines than the term suggests. A robot that walks factory floors and one that talks in a museum lobby are both humanoid. They are not the same investment, the same risk, or the same bet on the future.
0 Comments