
What happened
Parallel said GPT-6 Astra researched six labor-market statistics across four states over six months in half the time of prior models, with roughly 50% code cost reduction and the same research quality.
Why it matters
Doing the same research in half the time at roughly half the code cost means Parallel can divide work among multiple agents instead of running one long search sequence.
What to watch
The gains are measured on Parallel's own longest-running research tasks, so whether they hold for other workloads is undecided. Watch the 50% code cost reduction as the figure Parallel cites.
WHO IT HITSDevelopers building AI research agents — particularly teams doing long, multi-source research for financial and legal customers — could get the same quality answer with fewer research calls and less token spend, based on Parallel's test.
Summaries like this, every morning.
Parallel's business is developer infrastructure for AI agents that do knowledge work over the web, spanning web grounding for voice agents and research for financial institutions and legal customers. Its longest-running research tasks had typically required a bigger model with extended reasoning, which consumed more time and resources. The test with GPT-6 Astra focused on exactly that kind of task: an agent compiling six months of labor-market data across four states into a single report.
Beyond the headline time and cost figures, Parallel observed a change in how the agent searched. According to Devin Gupta, a Member of Technical Staff at Parallel Web Systems, Astra issued more targeted search queries and stayed focused on the ultimate task, incorporating its world knowledge compared with previous models. The company also noted that Astra can delegate specific research tasks to sub-agents, letting work happen simultaneously rather than moving through a single sequence of searches.
Taken together, these results point to a more practical way for Parallel to split demanding research across multiple agents without losing quality — though the reported improvements come from Parallel's own testing on its longest-running tasks, so how far they extend to other workloads is likely to depend on further results.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
OpenAI said Apollo Research, METR, Transluce and Goodfire will work on-site at its offices, with access to unr…

Walt Disney Imagineering R&D revealed three Audio-Animatronic prototypes—Vilo Vele, Kado Bre'a, and Jibu Bre'a…

Citibank changed its Online User Agreement this week so customers who grant third-party AI tools access to the…

Google rolled out temporary threads in AI Mode, letting users click an icon in the upper right corner to chat…

OpenAI CEO Sam Altman said at the New York Times DealBook Summit that he had spoken privately with former Nvid…

OneTrust launched its Fall '26 Release, built on OneTrust CORIE, adding a continuously updated AI inventory wi…