Resolves to YES if any model released in 2025 gets over 50% on the newly released GDPVal by April 1, 2026.
Resolves to NO if not.
Resolves to NO as early as January 1, 2026 if there is no plausible model that has not yet been evaluated, that might exceed 50%.
Update 2025-12-28 (PST) (AI summary of creator comment): - The scoring method will count wins+ties (not just wins), following OpenAI's official scoring methodology
Over 50% wins+ties will be sufficient for YES resolution
Creator plans to resolve YES on January 1 based on GPT-5.2-Pro results unless objections are raised
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ185 | |
| 2 | Ṁ135 | |
| 3 | Ṁ126 | |
| 4 | Ṁ98 | |
| 5 | Ṁ94 |
People are also trading
OpenAI's blog post and paper on GDPval for reference:
https://openai.com/index/gdpval/
https://cdn.openai.com/pdf/d5eb7428-c4e9-4a33-bd86-86dd4bcf12ce/GDPval.pdf
I'm slightly impressed that OpenAI created a new benchmark on which Claude beats them.
it's 2.4% away and it has now become a target