# Ginkgo Bioworks' Autonomous Laboratory Driven by OpenAI's GPT-5 Achieves 40% Improvement Over State-of-the-Art Scientific Benchmark

_Friday, June 26, 2026 at 11:33 PM EDT · science · Latest · Tier 2 — Notable_

![Ginkgo Bioworks' Autonomous Laboratory Driven by OpenAI's GPT-5 Achieves 40% Improvement Over State-of-the-Art Scientific Benchmark — Primary](https://mma.prnewswire.com/media/2878181/GinkosLab.jpg?p=facebook)

Ginkgo Bioworks announced that an artificial intelligence system it developed with OpenAI autonomously designed, executed and learned from biological experiments using the company's cloud laboratory infrastructure. The system combined OpenAI's GPT-5 reasoning model with Ginkgo's reconfigurable automation carts technology and Catalyst automation software.

In a preprint the collaborators reported that the GPT-5-driven autonomous lab reduced cell-free protein synthesis reaction costs by 40 percent relative to the state of the art. The system achieved production of superfolder green fluorescent protein at $422 per gram in total reaction component costs, compared with a previously reported $698 per gram.

The autonomous lab executed more than 36,000 experimental conditions in six rounds over six months. It ran more than 580 384-well plates and generated nearly 150,000 data points.

Human involvement was limited primarily to reagent preparation, loading and unloading, and system oversight. The model proposed and prioritized new reagents to test, some of which anticipated findings from published research.

Ginkgo said it is now selling the AI-improved reaction mix in its reagents store. The findings are described in a preprint available on OpenAI's website.

## Sources

- [PR Newswire](https://www.prnewswire.com/news-releases/ginkgo-bioworks-autonomous-laboratory-driven-by-openais-gpt-5-achieves-40-improvement-over-state-of-the-art-scientific-benchmark-302680619.html)

---
Canonical: https://techandbusiness.org/newswire/X0O85GNlLhBSz1ObTqvXXq
Retrieved: 2026-06-27T07:33:05.436Z
Publisher: Tech & Business (techandbusiness.org)