Seamless Communication
Accurately hears even in noisy environments, enabling natural dialogue
Service begins on the front lines
Kebbi 3
Next-Generation AI Partner
Connecting through understanding. Next-generation robot interaction.
Kebbi 3 is an AI communication robot specialized for frontline service operations. Powered by generative AI, it flexibly adapts to diverse needs and delivers natural, continuous conversation.
Key Features
Interactions with Deeper Mutual Understanding
Combines Edge AI and cloud AI to capture intent from visual and auditory information.
Integration with External Devices
Connects with external devices and systems, and can be configured to suit specific use cases.
Public information from third parties and government agencies
The following is public primary information confirming product listings, availability launches, and adopted technologies. It does not indicate deployment track records by end users.
Kanagawa Prefecture ROBOT TOWN SAGAMI
Featured on the Kebbi 3 product page as a communication robot combining generative AI and Edge AI.
現在、承認確認中Check the original sourceKanagawa Prefecture Sagami Robot Industry Special Zone
Kebbi 3 is listed as an eligible robot in the FY2026 robot adoption support subsidy materials.
現在、承認確認中Check the original sourceASRATEC Corp. Corp.
Announced the start of sales for Kebbi 3 manufactured by NUWA Robotics.
現在、承認確認中Check the original sourceMediaTek
Featured on the public Computex 2026 page, highlighting the adoption of MediaTek Genio 520 in Kebbi 3.
Check the original source
Link checked: 2026-08-28
Physical AI Working on the Front Lines
Embodied AI Agents Working on the Front Lines
Kebbi 3 does not keep generative AI responses confined to a screen. Combining voice, facial expressions, gestures, and touch interactions, it delivers services in the same space as people—from guidance, reception, and education to monitoring and business workflow execution.
EDGE AI × CLOUD AI
See, Hear, Understand, and Connect to the Next Action
Connects a 13MP camera, 4-microphone array, and 10 TOPS Edge AI with cloud AI and business data. Even in noisy environments, it distinguishes the speaker's voice, allowing you to design dialogue and task execution tailored to user situations and intent.
Open Integration
An Extensible Platform That Grows with Your Business
Connects to sensors, external devices, existing systems, and proprietary AI via USB 3.0 OTG, SDK, API, and Roflow. Start with a single customer service scenario and gradually expand features and locations based on operational data.
See Kebbi in Photos and Videos
Explore official NUWA materials featuring Kebbi product photos, live deployments in Japan, and BizTools demonstrations. Select a video card to play the full video on this page.
Official Product Introductions and Live Deployments in Japan
KebbiAlong with official Japanese introduction videos for Kebbi 3 and Kebbi, this collection features photos of real-world deployments in Japan—including corporate reception, transport roll-call checks, hotel front desks, showroom customer service, underground mall guidance, and expo exhibits.
1 / 20 · Video
Kebbi Demonstration Series
NUWA Official Channel4 videos: Face Editor, Content Editor, Programming Lab, and Conversation Training Kit. All are Japanese videos from our official channel. *Includes screens from older versions and Taiwanese audio. Some screens may differ from the latest version.
Features & System Overview
NUWA OfficialIncludes a demo of the generative AI dialogue at the core of Kebbi 3 (English audio and subtitles), an introduction video for the Robot Management System (RMS), and actual screens of the guidance UI used in live underground mall operations.
For detailed information about the tools, please visit the "NUWA BizTools" page.
Specifications
Main specifications
| Brain Core / SoC | MediaTek MTK G520（Edge AI 10 TOPS） |
|---|---|
| Memory / Storage | LPDDR5 8GB / eMMC 32GB |
| OS | Android 15 |
| Camera | 13MP (FOV 81°) / Infrared support |
| Connectivity | Wi-Fi 6 / Bluetooth 5.3 |
| Microphone | 4 digital microphone array (AIVC noise suppression, beamforming, sound source localization, voiceprint recognition) |
| Speaker | 3W speaker |
| Battery | 9,100mAh (Charging: 2-3h / Normal operation: 7-8h; varies depending on usage conditions) |
| Expansion | USB 3.0 OTG Type-C ×2 (Head, rear) / MicroSD card slot |
| Dimensions / Weight | 318 × 307 × 166 mm / 2.5kg |
| Motion / Input | 12 Servomotors (Body/Joints: 10 + Wheels: 2) / 7-inch LCD Touchscreen (1024×600) / 6 Touch Sensors |
Power / Sensors / Expansion
| Battery / Charging | 3.7V 9,100mAh Lithium-ion Battery / Charger Input: AC 100-240V 50/60Hz, Output: DC 5V-3A / DC 9V-2A (Rated 18W) |
|---|---|
| Sensors | PIR (Motion) Sensor ×1 / Ground Infrared Sensor (Drop Prevention) ×1 / Touch Sensors ×6 (Head, Abdomen, Palms) / Light Sensor |
| Lights / Indicators | RGB LED ×5 (Head, Hands, Body) / Listening Indicator / Status Indicator / Fill Light (Auxiliary Light) |
| Audio Output | Speaker 4Ω 3W |
| External Storage / Wired Connection | MicroSD Card (Supports up to 128GB) / Ethernet (Expansion via USB OTG) |
| Expansion Slot / Security | Ear Accessory Slot / Kensington Security Slot (K-Slot) |
Generation Comparison（Kebbi Air-S → Kebbi 3）
| Item | Kebbi Air-S | Kebbi 3 |
|---|---|---|
| AI Computing | CPU Only | 10 TOPS NPU+GPU+CPU |
| Memory | LPDDR3 3GB | LPDDR5 8GB |
| Audio | Basic Noise Reduction | AIVC + Sound Source Localization + Beamforming + Voiceprint |
| Camera | 5MP FOV63° | 13MP FOV81° |
| OS | Android 9 | Android 15 |
| Expansion | USB-C Charging | USB 3.0 OTG ×2 |
Frequently Asked Questions
What environments is Kebbi 3 suited for?
As an AI interaction terminal combining voice, screen, and facial expressions, it can be utilized in a wide variety of settings, including retail, healthcare and nursing care, education, hotel reception, and smart factories.
Can it integrate with generative AI and large language models?
Yes. It can integrate with LLMs, generative AI, NUWA Brain, and partner cloud AIs to serve as an interface for guidance, customer service, and consultations.
How has the hardware evolved from previous models?
It is equipped with 10 TOPS Edge AI, LPDDR5 8GB, a 13MP camera, a 4-microphone array, Android 15, and 2× USB 3.0 OTG. Recognition, response, and expandability have been enhanced for service environments.
What are the battery operating time and charging time?
Equipped with a 9,100 mAh battery, the estimated operating time during normal use is 7–8 hours, and charging takes approximately 2–3 hours. This varies depending on usage, such as continuous singing or dancing.
Source: NUWA Robotics Official Kebbi 3 Product InformationLast checked: 2026-08-24
Can it pick up voices even in noisy environments?
With a 4-digital-microphone array, AIVC noise suppression, beamforming, and sound source localization, it makes human voices easier to capture. We recommend pre-testing in your actual operating environment.
Can the wake word be changed?
In addition to the standard wake word, customization can be explored based on the deployment configuration and development method. Because the scope of support varies by project, please consult us via the inquiry form.
Is it waterproof and dustproof?
It is not waterproof or dustproof. Please avoid environments with water or dust that could cause malfunctions.
Is a network connection required?
While some pre-installed features operate locally, a Wi-Fi connection is required to fully utilize voice dialogue and AI capabilities.
Can we choose data storage locations and privacy measures?
Data flow and storage scope vary depending on the AI, cloud services, and integrated systems used. During implementation, we clarify collected data, storage locations, retention periods, and access privileges to propose a configuration that meets Japanese operational requirements.
Please tell me which features are available as standard and which require confirmation of terms of use.
Standard features include voice pickup via microphone array and smart noise canceling, conversations integrated with generative AI, movements and facial expressions powered by 12 servo motors, multilingual support, no-code task configuration via Roflow, and remote management via RMS. Voiceprint authentication, conversation interruption, custom wake words, generative AI model switching, and Roboslide require confirmation of availability and terms on a project-by-project basis. Long-term memory and multimodal AI interactions are currently under development, with release timing and terms subject to the manufacturer's official announcements. Please contact us via the inquiry form regarding implementation conditions.
Can software and hardware be customized for enterprise use?
Yes. Dialogue scenarios, screen interfaces, business system integrations, peripheral devices, and exterior accessories can be designed according to your requirements. Please reach out via the inquiry form for an initial consultation.
Where should I inquire about warranties and repairs?
Warranty terms in Japan vary depending on purchase and contract details. Please prepare the product name, serial number, symptoms, and purchase date, and contact us via the inquiry form.
Is there a place where I can see the actual robot?
現在、承認確認中At Nojima Corporation's robot showroom "MIRAI ROBO SQUARE" in Shinagawa, Tokyo, we exhibit a demonstration of Kebbi and Collibot collaborating in customer service. If you wish to visit, please contact us in advance and we will assist you after confirming exhibition availability.
We are considering adoption in the medical and nursing care fields. Where should we inquire?
現在、承認確認中Exclusively for the medical and nursing care fields, Marubun Corporation Corporation acts as the sales distributor. For other sectors or if you are unsure where to inquire, please contact us first via the inquiry form. We will review your inquiry and connect you with the appropriate representative.
See the differences from Collibot
Considering implementation?
Please feel free to contact us regarding implementation methods and our support system.