Details
- OpenAI says GPT-5.4 helped drive a medicinal chemistry project from literature review to a validated experimental result.
- The work was done with Maria AI and a specialized lab, which proposed a new way to improve Chan-Lam coupling, a reaction used to build drug-like molecules.
- The challenge focused on a difficult Chan-Lam variant involving primary sulfonamides, which has historically given low yields and limited usefulness in medicinal chemistry.
- GPT-5.4 reviewed scientific literature, generated and ranked research ideas, helped design experiments, analyzed results, and suggested follow-up studies.
- Human chemists guided the process, chose which proposals to test, and validated the final outcome.
- Maria tested the idea across 10,080 reactions, and human chemists later confirmed representative results by hand.
- Under the optimized conditions, yields improved for 88% of the boronic acids and 83% of the sulfonamides tested.
- OpenAI says the full process took about 2.5 months, plus another half month to write up the results.
Impact
The announcement suggests frontier models are moving beyond literature search into a larger share of the scientific workflow, including hypothesis generation and experiment planning. If replicated, that could speed iteration in medicinal chemistry and reduce the time needed to explore reaction conditions. It also reinforces a broader race in AI for science, where the competitive advantage is shifting toward systems that can connect models, lab automation, and human validation rather than just generate text.
