Agent Arena places DeepSeek-V4.1-Flash (Max) among its top open models at a reported $0.07 median cost per task

Agent Arena reports that DeepSeek-V4.1-Flash (Max) ranked third among open models with a 4.87% net improvement and a $0.07 median cost per task. The supplied comparison says this was near Hy4 preview’s reported improvement at 68% lower cost and below Kimi K3 (Max) at 91% lower cost. The excerpts do not provide the evaluation protocol or independently verify the leaderboard.