OpenAI has launched a dedicated science team to explore how its large language models can accelerate research. The move signals a strategic shift toward empowering scientists with AI tools that can process vast amounts of academic literature and suggest novel research paths.

The firm’s GPT-5 model has shown improved performance on advanced reasoning tasks, including solving graduate-level physics problems and scoring 92% on the GPQA benchmark. While earlier claims of solving unsolved math problems were overstated, the model still aids researchers by uncovering forgotten solutions and cross-disciplinary insights.

Compared to Google DeepMind, which has long focused on AI for science, OpenAI is a late entrant. However, executives believe that even incremental gains in research speed matter. Scientists are already using GPT-5 to sketch proofs, test hypotheses, and explore interdisciplinary connections at scale.

Industry watchers will monitor how OpenAI balances scientific ambition with practical utility. The real test lies in whether labs adopt these tools widely and if breakthrough discoveries emerge from human-AI collaboration.