Reaction
OpenAI models occupy 7 of 11 spots on Agent Arena's Pareto frontier
Agent Arena says GPT-6 Luna takes three of four optimal spots below $0.10 per task at three reasoning levels.
TLDR
Across different reasoning settings, GPT-6 Luna and GPT-6.1 Sol hold six of the 11 spots on Agent Arena's Pareto frontier, Agent Arena says. Luna claims three of four optimal spots below $0.10 per task; Sol sets the frontier from $0.22 to $0.65 per task. Adding GPT-6 Astra (Max) brings OpenAI models to seven of 11 spots. Agent Arena says changing a model's reasoning effort, rather than switching models, can sometimes reach an optimal cost-performance point.
Combined views
22.3K
2 Sources, first seen ago
