Google DeepMind’s Gemini 3.7 Flash has climbed to 20th place on the Agent Arena leaderboard, which evaluates AI models on real-world task performance. The model posted a 3.4% score improvement, outperforming its predecessors including Gemini 3.5 Flash (27th, +0.39%) and Gemini 3.6 Flash (around 34th). The ranking draws from over 1.7 million user sessions and tracks metrics such as success rates and tool hallucination. As of mid-August 2026, Gemini 3.7 Flash has not been officially released. Nonetheless, 19 models currently perform better on real-world agent tasks according to this ranking system.
Source: Read the original article
