DYNA Robotics launches DYNA 2.1, a physical agent that can finish a laundry shift

DYNA Robotics said Tuesday it launched DYNA 2.1, a semi-humanoid physical agent that can autonomously complete full commercial workflows — including an end-to-end laundry shift of loading, drying, folding, and shelving — and is being deployed in hotels, laundromats, and restaurants.

Most robot demos still win on a single tidy task. DYNA is selling shift-length commercial work, measured by how long the machine runs before a human has to step in, starting with laundry floors that hotels, laundromats, and restaurants already run.

On Tuesday, 29 September 2026, DYNA Robotics launched the DYNA 2.1 physical agent. The company calls it a semi-humanoid robot, powered by DYNA’s own model, that can finish an entire physical workflow on its own. A physical agent, here, is a robot that sees a room and carries a job through, not only a program that answers on a screen. Semi-humanoid means a torso and arms, on wheels, rather than a machine that walks on two legs. The release says the work needs both a careful grip and the ability to reason through a complicated job. The example it leads with is a commercial laundry shift. DYNA 2.1 is being deployed in hotels, laundromats, and restaurants. The PR Newswire page stamps the release Sep 29, 2026, 08:00 ET, which is 8:00 a.m. Eastern. The dateline is Redwood City, California. Those lines are DYNA’s.

The hardware, as the release describes it, is an upper torso with two arms. The arms work with parallel-jaw grippers or dexterous hands, so the robot can reach an object and handle it. A parallel-jaw gripper is a two-sided clamp. A dexterous hand has more fingers, closer to a person’s grip. The release prints the second option as “dextrous hands.” Four steerable wheels move the robot. DYNA says that base can get around without the risk of toppling. Those lines are DYNA’s. The release does not describe legs or a walking step.

A shift that runs for hours needs more than one clever grab. DYNA says the robot uses a vision-language orchestrator and a whole-body controller. The orchestrator looks at the scene and works through the workflow, the sequence of steps in the job. The controller lines up driving, reaching, bending, and lifting into one motion. Under that sits DYNA’s world action model, DYNA 2. A world action model is a system trained to pick the next physical move. DYNA says DYNA 2 was trained on a million hours of human and robot data, a million hours of recorded work, so it can handle physical skills and stay steerable. Steerable, here, means a person can still direct it. Those lines are DYNA’s.

The release says DYNA 2.1 can run a whole commercial workflow from start to finish. The examples it names are food preparation, data center management, and commercial laundry. Those lines are DYNA’s. The release does not name a hotel, a restaurant, a laundromat, or a data center as a customer.

In a commercial laundry, DYNA says the robot can finish an entire shift. It can load the washer and the dryer, and turn the machines on. It can bend to the back of a dryer and unload a pile of towels into a basket. It can snap towels flat, fold them, and sort them into stacks by size. It can crouch to put a finished stack on a low shelf, or reach up to put one on a high shelf. If it drops a towel, it picks the towel up. If a grasp goes wrong, it lets go and tries again. It does that without asking a person for help. Those lines are DYNA’s.

DYNA says it optimizes the robot for Mean Time Between Interventions, shortened to MTBI. That is how long the machine runs before a human has to step in. While it works, the robot records what it did, from the reasoner’s decision traces, the record of why it chose each step, to the controller’s joint loads, how hard each joint was working. That record feeds back into the system. DYNA calls the loop a continuous improvement loop, meant to raise the minutes of work the robot finishes between times a person has to help. Those lines are DYNA’s.

Lindon Gao, co-founder and chief executive, said: “Our goal is to make general purpose robots commercially viable for real-world scenarios.” A general-purpose robot, in that sentence, is one machine for many jobs, not a fixture bolted to one task. He said: “Our customers need robots that complete an entire workflow for a full shift – without human babysitters – not just specific tasks.” Those quotations are his, in the release. Jason Ma, co-founder, said a measure of how long the robot runs before a person steps in “is a far more useful benchmark than per-episode success rates on isolated tasks, which is widely used today but fails to reflect true commercial viability.” A per-episode success rate is the score for one tidy task, done once. He said: “Continuous real-world shifts require thousands of sequential steps, which is why we measure MTBI to ensure our robots are commercially viable and can operate without human babysitters.” Those quotations are his, in the release.

The about box says DYNA Robotics makes general-purpose robots powered by its own embodied AI foundation model. Embodied, here, means the model is tied to a body that acts in a room. DYNA says the model carries what it learns from one setting into another, and improves itself, at what the company calls commercial-grade performance. It says its robots have been deployed at customers across several industries. The founders are Lindon Gao and York Yang, who sold Caper AI for $350 million, and Jason Ma, a former DeepMind research scientist. The company is backed by investors including CRV and First Round. Those lines are DYNA’s. The release does not announce a new funding round, a price for the robot, or a customer’s name. The $350 million figure is the sale price of Caper AI, the founders’ earlier company. It is not a round for DYNA.

In plain terms, DYNA said on Tuesday that a wheeled robot with a torso and two arms can run a commercial laundry shift on its own: load the machines, turn them on, unload, fold, and shelve, and recover when a towel drops. The company says that machine is going into hotels, laundromats, and restaurants, and that the same kind of end-to-end work can cover food preparation and data center chores. It measures the job by how long the robot runs before a person has to help. The release does not name the customers or print a price.

The picture is a still from DYNA’s launch demo of that laundry shift, the video on the release. The robot stands at a row of stainless commercial washers, with one arm at an open round door, loading laundry. A screen beside it shows the scene the robot is using. It is a frame from the demonstration. It is not a photograph of a named hotel’s laundry room.