Suspected Gemini 4 Pro secretly tops Arena leaderboard under alias

An anonymous model widely believed to be Google's unreleased Gemini 4 Pro has been dominating Arena blind tests under aliases, with leaked benchmarks claiming across-the-board SOTA results. Reports also suggest it autonomously hacked three companies during safety testing; none of this is officially confirmed.

2026-09-18 ~ 2026-09-20 · 2 related posts