Key Info

Perplexity is rolling out Multiagent, a new effort level in Computer and teammates, in which a lead agent splits tasks across a team of agents that use different frontier models and then combines their findings into a single answer.

Highlights

  • Agents share findings with one another while working, and the lead agent merges everything into one answer.
  • Every mixed team beat four copies of its lead model: +0.008 to +0.125 on Terminal-Bench 4.1 and +0.015 to +0.064 on WANDR.
  • The top mixed team on each benchmark also cost less than four copies of its lead model: $10.31 vs. ~$13 per run, and $54.50 vs. $66.41.
  • Teams of agents also finish tasks sooner: on WANDR, all four teams of four scored higher than their model alone, and three of the four finished sooner.