Key Usage Statistics
OpenAI reports that roughly 8.4 million messages are exchanged each week about advanced science and mathematics, generated by about 1.3 million users worldwide.
This volume represents a 50% increase over the previous year, indicating that AI tools are moving from occasional experimentation to routine research workflows.
- ~8.4 M weekly messages on scientific topics
- ~1.3 M active users
- 50% year‑over‑year growth
Benchmark Achievements
The GPT‑5.2 series has delivered notable results on several high‑profile assessments.
- Gold‑level performance at the 2025 International Mathematical Olympiad
- Partial success on the FrontierMath benchmark
- 92% accuracy on the graduate‑level GPQA exam without external tools
Real‑World Applications
Researchers are integrating AI into everyday scientific tasks across multiple domains.
- Physics: AI assists in simulation integration, experimental log management, and theoretical exploration.
- Chemistry & Biology: Hybrid pipelines combine language models with graph neural networks and protein‑structure predictors, speeding up drug design and protein engineering.
- Mathematics: AI helps recombine known ideas, identify cross‑field connections, and accelerate formal verification and proof discovery.
Case studies include RetroBioSciences, where AI reduced protein‑design timelines from years to months.
Challenges and Future Outlook
Despite impressive usage figures and benchmark scores, independent validation remains limited.
Open questions include the durability of performance over time, the breadth of applicability across disciplines, and whether AI‑driven efficiencies translate into lasting scientific breakthroughs.