At a glance: Claude Opus 4.7 performs complex robotics tasks without human assistance 37 times faster than human teams from a year earlier and writes code that works correctly on the first attempt in most cases.
Anthropic has conducted “Project Fetch Phase Two” and shows that Claude Opus 4.7, without human support, manages complex robotics scenarios at significantly higher speed than human teams from the previous year. When controlling a robot dog, the new model version was on average 37 times faster than teams without AI support.
In August 2024, Anthropic conducted the original experiment with two teams of employees who, without robotics experience, were tasked with controlling a commercial robot dog. One team had access to Claude Opus 4.1, the other had to rely on the internet and their own initiative. The Claude team completed the tasks faster and more productively.
For Phase Two of the experiment, Anthropic revisited the same tasks with Claude Opus 4.7 in autonomous mode — meaning no human took physical control. The robot was supposed to execute the following steps in sequence: connect to video and lidar sensors, write a program for manual control, monitor the robot’s path, detect a beach ball, and finally retrieve it autonomously. Opus 4.7 required at least ten times less time than the fastest human team from 2024 for all completed tasks. For the four tasks that both human teams completed back then, Opus 4.7 was on average 37 times faster than the team without Claude and 18 times faster than the team with Claude 4.1.