共有:

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.

AI Navigate Editorial·2026.06.20·6 min read

Human 1x Claude 20x 20x faster Project Fetch · VentureBeat

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

Human Claude fastest team Opus 4.7 Human: baseline Claude: 20x faster
Project Fetch: task speed comparison

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