Gemini 3.8 Flash Reshapes Agent Arena's Pareto Frontier

Google DeepMind released Gemini 3.8 Flash (High), topping Claude Opus 5 on Text Arena and becoming the first model on LMArena's new Agent Arena Pareto frontier at just $0.22 per task. The Agent Arena leaderboard dynamically ranks 57 models across over 2.17 million real agent sessions.

2026-09-03 ~ 2026-09-03 · 4 related posts