Parallel cuts research time and cost in half using GPT‑6 Astra
Parallel, a developer‑infrastructure provider for AI agents that perform knowledge work over the web, says its new GPT‑6 Astra model cut research time by half and lowered code‑execution cost by roughly 50 % compared with earlier models. In a benchmark the company asked an agent to collect six labor‑market statistics across four states over a six‑month span, requiring the system to search…
Key points
- Parallel’s agent completed six labor‑market statistics across four states in half the time of prior models.
- The test showed roughly 50 % reduction in code cost while keeping research quality unchanged.
- GPT‑6 Astra enabled more targeted searches, fewer steps, and sub‑agent delegation for simultaneous work.
The test showed GPT‑6 Astra issued more targeted search queries, required fewer interaction steps, and could delegate sub‑tasks to subsidiary agents so that several searches ran in parallel. Parallel notes that the quality of the final report matched that of prior runs while the resource usage dropped, giving the platform a clearer path to scale demanding research workloads for financial, legal and other enterprise customers. The improvement also makes it practical to split complex questions among multiple agents, reducing waiting time and expanding the range of tasks that can be handled at scale.
Model page: GPT-6 Astra →
Parallel cut research time and cost in half with GPT‐6 Astra
OpenAI · 22 September 2026
Loading the full article…
This text was published by OpenAI. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
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.
More in Agents & Tools
All →- Opinion: chatgpt skills and plugins let users build personal AI assistants · 1 src
- OpenAI adds improved prompt caching to GPT-6, offering up to 90% input discounts · 1 src
- Opinion: Chris Green outlines key data sources for AI search · 1 src
- Meta says Muse heavily inspired by open-source OpenClaw · 1 src
- Microsoft adds AI video upscaling to Clipchamp on Windows 11 · 1 src
Comments
via GitHub Discussions