OpenAI: Near-Autonomous AI Chemist Based on GPT-5.4 Improves Reaction in Drug Synthesis

OpenAI, in collaboration with Molecule.one, presented a near-autonomous AI chemist based on GPT-5.4 that improved a key reaction in drug synthesis.

Reaction Optimisation in Medicinal Chemistry

The system near-autonomously improved a reaction in medicinal chemistry — the branch of chemistry concerned with the design and synthesis of drug compounds. “Near-autonomous” means that the AI leads the optimisation process with very few human steps, unlike the classical approach where chemists manually trial reaction conditions. The advance lies in the AI taking over part of the experimental decision-making rather than merely predicting outcomes.

A Step Towards AI-Driven Pharmaceutical Development

The demonstration is a step towards AI-driven research and development in pharmaceuticals, where synthesis optimisation is normally time-consuming and costly. Together with Google’s simultaneous publication of the AMIE medical AI, a clear direction is emerging: large models are increasingly entering real scientific and clinical workflows, not just text-based tasks.

What Should Be Taken with Caution?

The full publication returned a 403 error, so this account is based on the official OpenAI description. Specific details — which reaction was involved, how large the yield gain was, and how many human steps were removed — were not available in the source, and are deliberately not stated here to avoid speculation.