Resolves yes after GPT-5 is first benchmarked on IMO-2025. OpenAI's own reporting counts. Also resolves yes if the model achieves Silver or Gold.
Update 2025-07-25 (PST) (AI summary of creator comment): The creator has clarified the conditions under which the model's performance will be evaluated:
No scaffolds are permitted.
The model must be prompted with the questions exactly as written.
The model must not have access to tools or the internet.
1,000
3.00Seems largely dependent on whether this market permits custom scaffolds by external researchers. If Gemini 2.5 Pro could win Gold with custom elicitation, then GPT-5 could likely get at least Bronze. https://arxiv.org/abs/2507.15855
@bh For this market I'll say no scaffolds. Model must simply be prompted with the questions exactly as written with no tools are internet access