Parallel Search tops Artificial Analysis Search Index at 83
The jump from 75 to 83 in three months shows search APIs are improving fast enough to change agent cost models, not just accuracy rankings. Developers wiring web_search into agent harnesses should re-benchmark their current provider against Parallel's fast mode, which matches near-top accuracy at less than a quarter of the board's median cost per task. Against Exa and Perplexity, Parallel now leads on both quality and price, a combination worth testing before committing to annual API spend.