DigestAI news desk

Cut through the AI noise.

Agents & Tools3 min read

Perplexity adds live web searches to Decisions API for real-time answers

Perplexity launched live web grounding in its Decisions API on November 4, 2026, letting developers fetch real-time data before the model returns structured probabilities. The feature searches up to 10 URLs per question—limited to one second per search—before appending results as evidence. It targets dynamic decisions like executive roles, software updates, or event statuses, where answers may…

1 source

Key points

  • Live web grounding fetches up to 10 URLs per question before model answers, adding real-time evidence to probability outputs
  • Perplexity reports 97.3% accuracy on 328 claims with grounding, up from 56.4%, but latency rises to 366 ms median
  • Searches cost $0.001 each, capped at $0.10 for 100 questions, with retrieved data counted toward token limits

Perplexity claims accuracy jumped from 56.4% to 97.3% on 328 true/false claims when using live grounding, though latency nearly tripled (median 366 ms vs. 139 ms). On Parallel Web Systems’ benchmark, the grounded model achieved 86% accuracy on 100 public questions, surpassing its own previous score. Searches cost $0.001 each, with a $0.10 cap for 100 questions, but retrieved evidence also counts toward token limits. The API supports Python 0.44.0+ and TypeScript 0.39.0+, though type definitions for search results are still pending.

Full story from Unite.AI · by Aiden Cross, AI Product Strategy & Execution, AI Research AgentOpen source ↗

Perplexity Adds Live Web Grounding to Its Decisions API

Unite.AI · 4 November 2026

Loading the full article…

This text was published by Unite.AI and written by Aiden Cross, AI Product Strategy & Execution, AI Research Agent. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Topics · follow one to build your own front page
PerplexityParallel Web Systemspplx-decider-v1.2pplx-decider-flash-v1.2

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Agents & Tools

All →

Related stories