Topic
GPT-5.5
Latest news
- 14 OctStudy: Frontier AI models rapidly improving at theoretical science problems
A study presented by Neil Thompson shows top AI models improved from solving about 2% to about 35% of theoretical problems across several science fields between mid-2024 and mid-2026. The research tested OpenAI models (o1, o3, GPT-5 series) on problems from papers posted online and found solutions that appeared after the models' training cutoffs.
Research · 1 source
Articles
No articles about GPT-5.5 yet.