The Information reported that OpenAI solved 10 previously unsolved math problems with their soon-to-be released "Astra" model family. https://www.theinformation.com/briefings/exclusive-openai-previews-astra-ai-model-dc

I asked Claude Opus 5 (without internet) to name the 20 most important unsolved math problems, and it named the Jacobian Conjecture the 20th most important with the following list above it:

1. Riemann Hypothesis

2. Birch and Swinnerton-Dyer Conjecture

3. Langlands Functoriality

4. abc Conjecture

5. Twin Prime Conjecture

6. Goldbach's Conjecture

7. Odd Perfect Numbers

8. Schanuel's Conjecture

9. P vs NP

10. Unique Games Conjecture

11. Hadwiger's Conjecture

12. Collatz Conjecture

13. Navier–Stokes Existence and Smoothness

14. Invariant Subspace Problem

15. Kakeya and Restriction Conjectures

16. Hodge Conjecture

17. Yang–Mills Existence and Mass Gap

18. Smooth 4-Dimensional Poincaré Conjecture

19. Sphere Packing in General Dimensions

Resolves yes if OpenAI announces that Astra solved one of these problems. Resolves no if OpenAI does not claim that the model(s) expected to be called Astra solved any of the problems within a month of their announcement.

Only the first iteration of Astra models count. So if the first batch of announced Astra models don't solve one but version 2 does, it will not resolve yes.