Search is one of the most important tools for agents, and they search very differently to humans. By the end of the year, agents are expected to search the web far more than humans do
The Artificial Analysis Search Index measures the performance, cost and latency of Search APIs across 1700 tasks and three search-specific benchmarks: BrowseComp, DeepSearchQA and AA-Omniscience using the same base model and harness.
Developers can use the Artificial Analysis Search Index to inform their choices when it comes to what search tool their agent should be using. Now you can see who’s #1 while waiting for the bus. Currently Exa (auto) and Perplexity (medium) lead in performances with their best models scoring 80.