Particle.news
Download on the App Store
7 ARTICLES
·
3h ago

Parag Agrawal’s Parallel Unveils Deep Research API and Ultra8x Engine Outperforming GPT-5

The company says its flagship Ultra8x engine outperforms GPT-5 on key benchmarks, powering millions of daily AI research tasks.

Image
Agrawal, who began his career as a Twitter engineer before rising to CEO in 2021, kept a low profile after being removed when Musk.
Parag Agrawal’s New AI Firm ‘Parallel’ Outperforms GPT-5 and Humans in Web Research | Image: Twitter
Image

Overview

  • Parallel Web Systems, founded in 2023 and based in Palo Alto, has grown to a 25-member team after raising about $30 million from Khosla Ventures, Index Ventures and First Round Capital.
  • The Ultra8x engine and Deep Research API claimed 58 percent accuracy on the BrowseComp benchmark versus GPT-5’s 41 percent and a 25 percent human baseline, and secured an 82 percent win rate on DeepResearch Bench against GPT-5’s 66 percent.
  • The platform is built on eight specialized research engines and offers three developer APIs, including a low-latency option for chatbots and integrations for coding agents and enterprise workflows.
  • According to company announcements, Parallel already powers millions of AI-driven research tasks daily for startups and public enterprises, automating workflows from document analysis to market monitoring.
  • Positioning itself as a ‘web for machines,’ Parallel aims to let AI agents fetch, verify and organise live web data at scale rather than relying on static training sets.