Net Good IndexSubmit a correction

Capability Ledger · verified · Mathematics

AlphaProof and AlphaGeometry 2 reach silver-medal standard on IMO 2024

DeepMind’s AlphaProof and AlphaGeometry 2 solved four of six IMO 2024 problems for 28 of 42 points, scored by Timothy Gowers and Joseph Myers, one point below that year’s gold cutoff.

Tier 3 MajorMethodology 0.2
Submit a correction

Capability score

1.98

Does not count toward Net Good or the leaderboard.

10 base · Major (tier 3 of 5, 10 pts)
× 0.9000 attribution · Primary causal contribution
× 0.5500 evidence · Peer review or substantial independent validation
× 0.5000 realization · Complete public manuscript
× 0.8000 durability · Useful for many years
Event-level product before credit split: 1.98

First widely documented medal-level IMO performance by AI is a major capability milestone (tier 3). It is not Net Good: the problems were contest exercises, not open research. High attribution. Evidence between technical disclosure and later peer review of AlphaProof (0.55).

What happened

On 25 July 2024, DeepMind reported that AlphaProof (Lean reinforcement learning) solved IMO 2024 problems 1, 2, and 6, and AlphaGeometry 2 solved problem 4. Combinatorics problems 3 and 5 were unsolved. Formalization of statements into Lean was done by humans; computation ran over days, not contest time. A 2025 Nature paper later described AlphaProof. The medal analogy is unofficial: the systems were not contestants. This is a capability milestone on a contest paper, not a contribution of new theorems, and does not count toward Net Good.

Model attribution

1.39

AlphaProof

Found Lean proofs for three non-geometry IMO 2024 problems, including P6.

Solved three of the four solved problems, including the hardest.

Attribution 0.9000 · Credit share 70% · Google DeepMind

0.59

AlphaGeometry

Proved the IMO 2024 geometry problem.

Solved the single geometry problem in the four-problem total.

Attribution 0.9000 · Credit share 30% · Google DeepMind

Claims

  • The combined systems solved four of six IMO 2024 problems for 28 points as scored by Gowers and Myers.

    outcome · supported

  • The systems competed as official IMO contestants under contest time limits.

    attribution · disputed

Sources

primary sources

independent sources

Secondary domains: Computer Science

Revision history

  • 13 Sep 2026 · 0.00 → 1.98

    Initial adjudicated seed score under methodology 0.1.

  • 14 Sep 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 17 Sep 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 18 Sep 2026 · 1.98 → 1.98

    Ingest update: added a follow-up source.

  • 18 Sep 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 19 Sep 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 21 Sep 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 6 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 7 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 11 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 11 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 21 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 24 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 27 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

  • 28 Oct 2026 · 1.98 → 1.98

    Methodology 0.2: contest-paper performance moved to the Capability Ledger; not Net Good.

Submit a correction