Anthropic Releases Phase 2 of Project Fetch: Claude Opus 4.7 Achieves 10x Speed Improvement in Robot Tasks

Odaily, February 6 - Anthropic has announced the results of Phase 2 of "Project Fetch," evaluating the performance improvements of its latest model in real-world robotic operations. Conducted in August 2025, the experiment tasked Anthropic employees—who are not robotics experts—with completing a series of complex missions using off-the-shelf quadruped robots. The study compared the performance of "teams assisted by Claude models" against those relying solely on "humans and the internet." The results showed that when operating fully autonomously with the latest Claude Opus 4.7 model, it significantly outperformed human teams in average speed across all achievable tasks, completing operations at least 10 times faster.

Anthropic stated that this progress is not the result of specialized robotic training but rather an extension of general large model capabilities, suggesting that AI is moving from "assisting humans in using tools" toward an early phase of "directly operating physical tools." This mirrors the evolutionary path previously observed in software engineering, moving towards agentic programming.