Gemini 4 Carbon debuts at No. 1 on APEX-Agents leaderboard
Gemini 4 Carbon scored 84.2% Pass@1, leading consulting and banking tasks but placing second in corporate law.
TLDR
Mercor reports that Carbon earned a 90.4% mean on APEX-Agents, the only model on the leaderboard above 90%. The benchmark tests multi-step tasks in management consulting, investment banking and corporate law. Gemini 4 Argon placed second overall but led corporate law with 76.7% Pass@1, compared with Carbon’s 73.8%.
Combined views
8.4K
3 Sources, first seen ago

