An update to my earlier post on Perplexity Computer running fully local on a Mac. That launch was about keeping the work on the device. This one is a different mode. Perplexity has added Multiagent, a new effort level above Ultra, MarkTechPost reports, following the Light, Standard, High, and Ultra controls that arrived on September 17, 2026. A lead agent can assign up to 16 supporting agents, plus take on part of the work itself. Those agents run on different frontier models, work in parallel, share findings directly, and check one another’s work, instead of only reporting back to the lead.
In Perplexity’s example, Claude Fable 5.2 plans the task and splits it among agents running Claude Opus 5.5, GPT 6.1 Sol, and Gemini 4 Carbon, the default team. Fable 5.2 and GPT 6 Astra join as supporting agents only when requested or needed. Users still talk to the lead, which merges the result. You can name models in the query, see which agents are running, and stop the whole team by stopping the thread. MarkTechPost says Multiagent is available on the web for Perplexity Max subscribers, with no local or quantized option. That is not the on-device Mac mode I wrote about earlier, and Perplexity has not published accuracy or speed numbers for the new level.
My take
The useful distinction is the one Perplexity is drawing between Ultra, Multiagent, and Model Council: one lead with helpers, a team that talks to each other, or several models answering the same question independently. I have not run Multiagent, and I would not confuse it with the local Mac computer. If you are already on Max and have a messy, many-part research job, this is the new slider position to try. If you wanted the work to stay on the machine, this is not that feature.
God Bless, and May the Source Be with You!