Parallel cut research time and cost in half with GPT‑6 Astra

Parallel⁠(opens in a new window) builds developer infrastructure for AI agents that do knowledge work over the web. Its tools support everything from web grounding for voice agents to research for financial institutions and legal customers, combining frontier models with web search.

For Parallel’s longest-running research tasks, getting a high-quality answer typically meant using a bigger model with extended reasoning, which consumed more time and resources. The company has seen a major improvement in time and cost with GPT‑6 Astra.

“With Astra, we’ve demonstrated that you can get the same high-quality research much, much faster with fewer research calls and less tokens.”

—Devin Gupta, Member of Technical Staff, Parallel Web Systems

Compiling six months of data twice as fast

In one test of GPT‑6 Astra, Parallel asked its agent to research six different labor-market statistics across four states over six months. The agent had to search across multiple websites, collect information, and compile the findings into a single research report.

GPT‑6 Astra was able to complete the work in half the time of prior models, with roughly 50% code cost reduction, while delivering the same quality of research.

Reaching high-quality answers in fewer steps

Parallel also observed that GPT‑6 Astra made more focused searches and took fewer steps to reach a useful result.

“Astra issued more targeted search queries and focused on the ultimate task better, incorporating its world knowledge compared to previous models.”

—Devin Gupta, Member of Technical Staff, Parallel Web Systems

The increased efficiency also makes it more practical for Parallel to divide research among multiple agents. GPT‑6 Astra can delegate specific research tasks to sub-agents, allowing work to happen simultaneously and reducing the time spent moving through a single sequence of searches.