From the source
Perplexity is rolling out Multiagent, a new effort level in Computer and teammates that has a lead agent split tasks across a team of agents running different frontier models.- As the agents work, they share findings with one another and the lead agent combines everything into a single answer.
- Perplexity says Multiagent is optimized for performance and is credit-intensive, so it should be used only for the most important tasks, such as forecasting, high-density problem solving and consequential decision making.
- On Terminal-Bench 4.1, Perplexity ran five frontier models alone, as two and four copies, and leading a mixed-model team of four; each mixed-model team outperformed both the lead model and multiple copies.
- The mixed team led by Gemini 4 Carbon scored highest, at 0.727.
- Perplexity says the results hold on both benchmarks, with every mixed team beating four copies of its lead model by +0.008 to +0.125 on Terminal-Bench 4.1 and +0.015 to +0.064 on WANDR.
- The top mixed team on each also cost less than four copies of its lead: $10.31 vs. ~$13 per run, and $54.50 vs. $66.41.
- Teams of agents also complete tasks sooner: on WANDR, every team of four scored higher than its model alone, and three of the four finished sooner.



