In a Nov. 4, 2026, post, Perplexity’s developer account, @perplexitydevs, announced that its Decisions API can search the web before answering. Developers enable web grounding with "mode": "web_grounding". Fast Search then runs one query per question.
The developer account explains that search results join the request’s state as evidence before the decision model answers. Developers can also pass a URL as the state to ask about a page. Perplexity suggests using web grounding to fact-check a list of claims, check for a newer stable release of a pinned package, or flag accounts with a new CEO, an acquisition or layoffs.
In a separate post, @AravSrinivas described the integration as allowing batch-mode decisions grounded in web information. The post lists search at $0.001 per request and Decisions API input at $0.02 per 1 million tokens. Those are separate rates, not an all-in price per decision.
Perplexity that its pplx-decider-v1.2 model with web grounding scored 86% on 100 public PWS-Decisions questions, compared with 51% without web grounding. In that test, the company reported median latency of 370 milliseconds and p95 latency of 685 milliseconds. The figures describe Perplexity’s test, not performance across all API requests.
