Agility's Digit 5 sheds the safety cage and the ostrich legs — the most-deployed US humanoid gears up for scale.

On September 15, Salem, Oregon-based Agility Robotics unveiled Digit 5, the next generation of its general-purpose humanoid and, the company says, its first engineered for 'cooperatively safe work at scale' — meaning it can work in close proximity to people without the physical safety barriers that traditional industrial automation requires. Agility says it built the robot to the exact requirements customers handed it after three years of the previous Digit 4 working on their production floors. That deployment record is what separates Digit from most of the field: it is one of the only humanoids actually doing paid work in manufacturing, distribution and logistics today, with more than 65,000 hours of operation across GXO, Schaeffler, Amazon and Toyota Motor Manufacturing Canada. At GXO's Flowery Branch facility near Atlanta, a Digit 4 hit a cumulative 100,000-tote milestone at roughly 98% accuracy.

The hardware changes are substantial and pointed squarely at real work. The most visible is the legs: Agility dropped the backward-bending, ostrich-like knees of earlier Digits for a more human configuration, powered by proprietary cycloidal actuators, that lets the robot repeatedly lift up to 50 pounds (22.7 kg) — 40% more than Digit 4 and enough, Agility says, to cover single-person lift tasks in OSHA-regulated facilities. A new 90-minute battery recharges in just nine minutes, a 10:1 run-to-charge ratio (up from Digit 4's 2:1) that Agility says enables more than 20 hours of productive work in a 24-hour day. Digit 5 stands 5 feet 11 inches (1.81 m), weighs 284 pounds (129 kg) and now reaches up to 7.2 feet (from 5.5), while still fitting the aisles, doorways and workstations built for people. Swappable grippers on ISO-standard mounting flanges let it move from tote handling to depalletizing, machine tending, kitting, quality inspection and palletizing, and the robot runs about 30 motors with motor-level control roughly 1,000 times a second.

The headline feature is really the safety architecture, because safety — not dexterity — has been the wall keeping humanoids roped off in cages away from workers. Digit 5 continuously watches for people using proprietary AI and multiple sensors across a 360-degree radius, and responds by slowing, stopping, stepping aside or sitting down and cutting its motors so a person can safely approach. It layers on visual and audible cues plus an independent safety controller as a final backstop. Agility says Digit was the first humanoid to pass an independent field evaluation for industrial safety administered by the U.S. Occupational Safety and Health Administration, and it is the first launch partner for NVIDIA's Halos for Robotics safety platform, built on NVIDIA IGX Thor. The company is also helping write the rules it will be judged by, contributing to the ANSI/A3 R15.108 technical report and the ISO 25785-1 standard, the first international safety standard for humanoids.

For a US buyer, Digit 5 is the most credible near-term option in the West: it is US-made (which sidesteps the FCC's July import ban on Chinese robots), it can work next to people, and it comes from the vendor with the longest real deployment track record. But the caveats are the ones that matter most. This is an order, not stock — early access does not begin until the first half of 2027, with general availability, plus first-time EU and UK sales under a CE mark, targeted for the end of that year. The reported $300 million in multi-year orders, described as of May 2026, is subject to contractual milestones, and it should be read against the reality that Agility booked just $1.8 million in net sales in 2025 per its SEC filings as it prepares to go public through a SPAC merger with Churchill Capital Corp XI at a $2.5 billion valuation. The gap between a $300 million order book and $1.8 million of revenue is the honest measure of how early this still is. If you are evaluating Digit 5, press for the autonomous success rate and cycle time on your specific tasks, how many robots one operator supervises, and dated multi-shift output at a named site before you plan a line around it.