Robots/Agility
Agility Digit
Humanoidworking for customers, price not disclosed
22 sourced facts: 22 confirmed. Last checked 14 Oct 2026.
- Confirmed
- The robot's maker or the part's maker says so.
- Likely
- A good source points to it, but neither maker has said it outright.
- Rumoured
- One unverified report. A lead, not a fact.
Point at any row to see its source.Tap any row to see its source.
Availability
Whether you can get one, and what it costs.
Working for customersDigit v4Confirmed
Digit is doing paid work for companies including Schaeffler, GXO and Toyota Motor Manufacturing Canada.
- Applies to
- Digit v4
- Also see
- NVIDIA blog, NVIDIA Jetson Thor Unlocks Real-Time Reasoning for General Robotics and Physical AI (Aug 25, 2025)blogs.nvidia.com
- Checked
- 14 Oct 2026
Open for ordersDigit 5Confirmed
The next model, Digit 5, already has more than 300 million dollars of multi-year orders. Early access is expected in the first half of 2027 and general availability by the end of 2027.
- Applies to
- Digit 5
- Source
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale (Sep 15, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Not disclosedAgility does not publish a price. Its 2026 merger filing uses about US$200,000 to buy one, or US$8,500 a month to rent plus about US$25,000 to set up, as planning figures rather than a quoted price.
Agility does not publish a price. Its 2026 merger filing uses about US$200,000 to buy one, or US$8,500 a month to rent plus about US$25,000 to set up, as planning figures rather than a quoted price.
- We looked at
- Churchill Capital Corp XI (merger partner), Form 424B3 proxy statement/prospectus, Illustrative Unit Economicssec.gov
- Also see
- Agility, FAQs (pricing)agilityrobotics.com
- Checked
- 14 Oct 2026
Brain layer 1 of 5
AI model
The software that decides what the robot does.
Agility whole-body control modelConfirmed
Agility's own small AI model that controls how Digit walks, balances and moves its whole body. Agility built its own small AI model that controls how Digit walks, balances and moves its arms and body, like a motor sense. It is learned entirely in simulation and then works on the real robot.
- Made by
- In-house
- Applies to
- All versions
- Source
- Agility, Training a Whole-Body Control Foundation Model (Aug 28, 2025)agilityrobotics.com
- Also see
- Agility, Agility and AI (Mar 16, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Gemini Roboticsresearch useConfirmed
Google DeepMind's family of AI models that help robots understand what they see, follow instructions and plan movements. Agility has tested steering Digit's movement model with a research preview of Google DeepMind's Gemini AI, and DeepMind lists Agility among early testers of its Gemini Robotics-ER model. This is described as research, not what runs on customer robots.
- Made by
- Google DeepMind
- Applies to
- research use
- Source
- Agility, Training a Whole-Body Control Foundation Model (Aug 28, 2025)agilityrobotics.com
- Also see
- Google DeepMind, Gemini Robotics brings AI into the physical world (Mar 12, 2025)deepmind.google
- Checked
- 14 Oct 2026
Brain layer 2 of 5
Chip
The computer inside the robot that runs it.
IGX ThorDigit 5Confirmed
An industrial Nvidia computer for robots, with built-in safety features. Digit 5 uses Nvidia's IGX Thor, an industrial computer with built-in safety features, to run its system that spots nearby people.
- Made by
- Nvidia
- Applies to
- Digit 5
- Source
- NVIDIA, NVIDIA Announces Halos for Robotics, the Industry's First Full-Stack Safety System for Physical AI (Jun 22, 2026)investor.nvidia.com
- Also see
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scaleagilityrobotics.com
- Checked
- 14 Oct 2026
Jetson ThorA future generation, plannedConfirmed
Nvidia's newest onboard AI computer, aimed at humanoid robots. Agility plans to use Nvidia's Jetson Thor as Digit's main onboard computer in a coming generation, so bigger AI models can run on the robot itself. This was announced as a plan, not as shipping hardware.
- Made by
- Nvidia
- Applies to
- A future generation, planned
- Source
- NVIDIA blog, NVIDIA Jetson Thor Unlocks Real-Time Reasoning for General Robotics and Physical AI (Aug 25, 2025)blogs.nvidia.com
- Also see
- Agility, Agility Robotics Powering the Future of Robotics with NVIDIA Jetson Thor (Aug 25, 2025)agilityrobotics.com
- Checked
- 14 Oct 2026
Jetson, model not namedCurrent DigitConfirmed
One of Nvidia's small computers for running AI inside machines. Which one has not been said. Nvidia says the current Digit has one of its Jetson computers inside, but does not say which one.
- Made by
- Nvidia
- Applies to
- Current Digit
- Source
- NVIDIA blog, NVIDIA Jetson Thor Unlocks Real-Time Reasoning for General Robotics and Physical AI (Aug 25, 2025)blogs.nvidia.com
- Also see
- Agility, Agility Robotics Expands Relationship with NVIDIA (Mar 18, 2025)agilityrobotics.com
- Checked
- 14 Oct 2026
Brain layer 3 of 5
Senses
How the robot sees and hears the world.
RealSense D455Confirmed
A depth camera that sees in 3D and works at longer range than its smaller sibling. A RealSense D455 depth camera sits in Digit's neck and gives it a 3D view for spotting objects, avoiding collisions and mapping the ground.
- Made by
- RealSense
- Applies to
- All versions
- Source
- RealSense, RealSense and Agility Robotics: Augmenting Human Potentialrealsenseai.com
- Checked
- 14 Oct 2026
Onboard camerasDigit 5Confirmed
Cameras built into the robot. Digit 5 uses several onboard cameras to find its way and work safely, and handles that data on the robot itself.
- Made by
- Not named
- Applies to
- Digit 5
- Source
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale (Sep 15, 2026)agilityrobotics.com
- Also see
- Agility, FAQsagilityrobotics.com
- Checked
- 14 Oct 2026
Depth cameraConfirmed
A camera that measures how far away each thing it sees is. Digit has infrared depth sensors that measure how far away things are, helping it notice unexpected obstacles and stop or go around them.
- Made by
- Not named
- Applies to
- All versions
- Source
- Agility, FAQsagilityrobotics.com
- Also see
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scaleagilityrobotics.com
- Checked
- 14 Oct 2026
3D lidarConfirmed
A laser scanner that measures distance in every direction. Digit carries lidar, a laser scanner that measures the space around it, alongside its cameras. The maker of the lidar is not named.
- Made by
- Not named
- Applies to
- All versions
- Source
- Agility, FAQsagilityrobotics.com
- Also see
- IEEE Spectrum, ROBOTS Guide: Digit (original version lists lidar and four Intel RealSense depth cameras)robotsguide.com
- Checked
- 14 Oct 2026
Motion sensor (IMU)Digit 5Confirmed
A sensor, like the one in a phone, that feels tilt and movement so the robot can keep its balance. Digit 5 has motion sensors that feel tilt and acceleration, which help it keep its balance.
- Made by
- Not named
- Applies to
- Digit 5
- Source
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale (Sep 15, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Force sensingConfirmed
Sensors that measure how hard the robot is pushing, pulling or gripping. Digit records force readings along with camera images and joint positions, so it can sense how hard it is pushing or gripping. The sensor maker is not named.
- Made by
- Not named
- Applies to
- All versions
- Source
- Agility, Agility and AI (Mar 16, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Brain layer 4 of 5
Training
How the robot was taught.
Isaac LabConfirmed
The newer version of Nvidia's robot training simulator. Agility trains Digit's skills first in Nvidia's Isaac Lab simulator, where a virtual robot practises by trial and reward.
- Made by
- Nvidia
- Applies to
- All versions
- Source
- Agility, Agility Robotics Powering the Future of Robotics with NVIDIA Jetson Thor (Aug 25, 2025)agilityrobotics.com
- Also see
- Agility, FAQsagilityrobotics.com
- Checked
- 14 Oct 2026
Isaac SimConfirmed
Nvidia's simulator for designing and testing robots, which also produces computer-made training images. Digit's body-control model was trained in Nvidia's Isaac Sim physics simulator, packing decades of practice into a few days.
- Made by
- Nvidia
- Applies to
- All versions
- Source
- Agility, Training a Whole-Body Control Foundation Model (Aug 28, 2025)agilityrobotics.com
- Also see
- NVIDIA, Customer story: Agility Robotics Digit humanoid robotnvidia.com
- Checked
- 14 Oct 2026
MuJoCoConfirmed
An open physics simulator used to test robot movement. Agility also uses MuJoCo, a physics simulator from Google DeepMind, to build virtual worlds for training Digit.
- Made by
- Google DeepMind
- Applies to
- All versions
- Source
- Agility, Agility and AI (Mar 16, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Reinforcement learningConfirmed
Learning by trial and error, keeping what works. Digit learns balance and handling skills by trial and error in simulation, getting a reward when it does well, before it works in a real building.
- Made by
- Not named
- Applies to
- All versions
- Source
- Agility, FAQsagilityrobotics.com
- Checked
- 14 Oct 2026
Teleoperated demonstrationsConfirmed
People drive the robot remotely, and the recordings become its lessons. Engineers wearing VR headsets remotely guide Digit through tasks many times, and those recordings are used to teach it.
- Made by
- Not named
- Applies to
- All versions
- Source
- Agility, Agility and AI (Mar 16, 2026)agilityrobotics.com
- Checked
- 14 Oct 2026
Motion captureConfirmed
Recording how real people move with tracking equipment, so a robot can learn from those movements. Agility also teaches Digit from recordings of real human movement, captured with motion-capture equipment.
- Made by
- Not named
- Applies to
- All versions
- Checked
- 14 Oct 2026
OmniverseConfirmed
Nvidia software for building realistic virtual copies of real places, used to test robots before they go to work. Agility tests trained skills in virtual copies of real facilities built with Nvidia Omniverse before using them on real robots.
- Made by
- Nvidia
- Applies to
- All versions
- Source
- NVIDIA, Customer story: Agility Robotics Digit humanoid robotnvidia.com
- Also see
- Agility, Agility Robotics Expands Relationship with NVIDIA (Mar 18, 2025)agilityrobotics.com
- Checked
- 14 Oct 2026
Brain layer 5 of 5
Software
The system that ties it together and lets people program it.
Agility ArcConfirmed
Agility's online software for setting up a fleet of Digit robots, giving them jobs and tracking their work. Agility Arc is the company's cloud software for setting up Digit robots, giving them jobs, watching how they perform and sending them updates.
- Made by
- In-house
- Applies to
- All versions
- Source
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale (Sep 15, 2026)agilityrobotics.com
- Also see
- Agility, Solutionsagilityrobotics.com
- Checked
- 14 Oct 2026
Halos CoreDigit 5Confirmed
Nvidia safety software that handles the safety-critical parts of a robot's system. Digit 5's people-detection safety system uses Nvidia's Halos Core safety software running on the IGX Thor computer.
- Made by
- Nvidia
- Applies to
- Digit 5
- Source
- NVIDIA, NVIDIA Announces Halos for Robotics, the Industry's First Full-Stack Safety System for Physical AI (Jun 22, 2026)investor.nvidia.com
- Also see
- Agility, Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scaleagilityrobotics.com
- Checked
- 14 Oct 2026
Not disclosedAgility does not publish a public software kit for outside programmers, and none of its pages say Digit runs ROS 2. Customers connect Digit to their warehouse systems through Agility Arc.
Agility does not publish a public software kit for outside programmers, and none of its pages say Digit runs ROS 2. Customers connect Digit to their warehouse systems through Agility Arc.
- We looked at
- Agility, FAQsagilityrobotics.com
- Checked
- 14 Oct 2026
Changes
None since this page was published on 14 Oct 2026. When a fact is corrected, the old version stays listed here.
Spotted a mistake?
Send a link that shows the right answer. Public sources only.