Buddy Tests AI Agents Controlling Real Android Phones

An open-source Android app called Buddy lets coding agents like Claude Code directly operate real phones. On a OnePlus 12 benchmark, Sonnet 5.5 topped the test, completing all tasks in 25 seconds with a 10/10 success rate.

2026-10-18 ~ 2026-10-18 · 2 related posts

• Open-source Buddy benchmark tests which AI model uses a real Android phone fastest — 3amae404 · 2026-10-18
• Open-source Buddy benchmark: Sonnet 5.5 fastest at driving a real Android phone at 25s — 3amae404 · 2026-10-18