tech
December 12, 2025
Advancing science and math with GPT-5.2
GPT‑5.2 is our strongest model yet for math and science work.
TL;DR
- GPT-5.2 Pro and GPT-5.2 Thinking are the strongest models yet for scientific and mathematical work.
- The models show improved performance on benchmarks like FrontierMath and GPQA Diamond.
- GPT-5.2 Pro assisted in resolving an open research problem in statistical learning theory regarding learning-curve monotonicity.
- The AI directly solved the problem, with human researchers verifying the proof and extending the results.
- AI systems can support scientific research by exploring proofs and testing hypotheses, but human judgment and verification remain essential.
Continue reading
the original article