Perplexity's research team published WANDR, a benchmark for evaluating research agents on high-volume, evidence-heavy knowledge work that requires both broad and deep searching. The posting, dated 14 July 2026 on research.perplexity.ai, frames the benchmark around tasks where an agent must gather and cite large volumes of evidence rather than answer a single lookup. Beyond the framing, the retrieved page contained only site navigation, so the benchmark's task counts, scoring method and model results are not available from it.