If you use Deep Research modes regularly, you know thorough data gathering takes patience. Kick off a full competitor analysis or financial data sweep, and you wait ten minutes while the agent crawls dozens of sites and drafts a report.
The Perplexity AI team went straight after the biggest bottleneck: inference speed. Inside the Perplexity Computer environment, they rolled out an UltraFast preset that runs background workflows five times faster.
As Julian Goldie SEO points out, the system switched to dynamic parallel query routing. Instead of walking pages one by one, the algorithm queries dozens of sources simultaneously, filters out dead ends on the fly, and assembles summary tables in one to two minutes.
There's a predictable trade-off: recursive depth on secondary links is slightly shallower. Still, for 90% of day-to-day analytical tasks on Pro and Enterprise plans, it hits a sweet spot between speed and report completeness.
