Perplexity launches Multiagent: diverse model teams beat homogeneous copies on benchmarks

AravSrinivas · x · 2026-11-18

Perplexity CEO Arav Srinivas announced Multiagent, usable with Computer and Teammates (toggle via Effort level). One model can orchestrate others in parallel agent streams to finish complex tasks faster and more accurately — pitched for financial research, forecasting, prediction markets, and batch filing analysis. Available to all Max users.

He claims model diversity matters: on both Terminal Bench and WANDR, diverse model teams outperform homogeneous copies of the same agent on both performance and cost — the differentiation of a multi-model platform.

Related event: Perplexity Launches Multiagent Effort Level for Multi-Model Parallel Collaboration(6 posts)→

Original post →

More from coding & agent

coding & agent channel →