Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data

Artificial Analysis has released Optima, a platform that allows users to create custom AI benchmarks using their own specific data and operational workflows. By evaluating models based on internal task performance, cost, and time, the tool provides a more practical assessment for businesses than generic industry tests. This approach is particularly relevant for agent-based applications, where efficiency and outcome quality are often more critical to the user than base token pricing.