Robotics · Claude (Anthropic)
Claude Opus 4.7 completes robotics tasks 20x faster than humans
Project Fetch: Claude Opus 4.7 clocked ~20x the speed of the fastest human team, unassisted.
Before
For the past few years, AI robotics has been stuck at the 'can write code but cannot move things' stage.
Language models got good at generating code and plans, but actual physical-space execution — recognizing, grasping, and transporting objects — lagged badly. Demo videos and real-world utility stayed far apart.
AI robotics benchmarks remained narrow lab metrics; head-to-head comparisons with humans showing a decisive gap were essentially absent.
What Changed
Claude Opus 4.7 completed a robotics challenge unaided ~20x faster than the fastest human team (Project Fetch, VentureBeat)
20x
vs best human team
Unaided
no intervention
Opus 4.7
model used
Impact for You
Physical task automation is catching up to real-world speeds. Precise manipulation still favors humans, so hold off on broad manufacturing rollouts.
This week, AI reached the battlefield and the robotics floor. Repetitive physical tasks — warehouse sorting, transport, retrieval — are now serious candidates for AI automation. Assembly and inspection work that demands fine motor precision still needs human oversight.
Source: Claude (Anthropic) official