AlphaProof and AlphaGeometry 2 reach IMO silver-medal standard

AI system

System: AlphaProof and AlphaGeometry 2
Model: AlphaProof and AlphaGeometry 2 in the 2024 work.
Release: Exact release not established.
Application developer: DeepMind
Role: The systems generated mathematical proofs.
Configuration: The 2024 work combined AlphaProof and AlphaGeometry 2. A later report concerned Gemini Deep Think.
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Manual translation of the 2024 problem statements into Lean. AlphaProof reinforcement-learning proof search and AlphaGeometry 2 for the geometry problem. Grading under IMO rules by two former IMO gold medalists.
Human oversight: Not assessed
Relation to AI: The AI systems produced the proofs. Humans manually formalized the 2024 problem statements into Lean (Scientific American) and graded the output. In 2024 the solutions were graded by two mathematicians at DeepMind's request (one of them, Myers, chaired the IMO 2024 Problem Selection Committee). The result was not officially certified by the IMO. DeepMind states the 2025 result was graded and certified by IMO coordinators, with the IMO president's confirmation appearing in the vendor blog. The cited sources do not state whether OpenAI's 2025 claim was certified by the IMO.
AI contribution: The AI systems produced the proofs. Humans manually formalized the 2024 problem statements into Lean (Scientific American) and graded the output. In 2024 the solutions were graded by two mathematicians at DeepMind's request (one of them, Myers, chaired the IMO 2024 Problem Selection Committee). The result was not officially certified by the IMO. DeepMind states the 2025 result was graded and certified by IMO coordinators, with the IMO president's confirmation appearing in the vendor blog. The cited sources do not state whether OpenAI's 2025 claim was certified by the IMO.

Benefits

Outcome identifiers: Not assessed
Benefit: awarded 28 of 42 points, a silver-medal score
Category: Improved Task Accuracy
Status: Not assessed
Beneficiary groups: Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: awarded 28 of 42 points, a silver-medal score
Comparison: Not assessed
Results: awarded 28 of 42 points, a silver-medal score
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: DeepMind. Operator: DeepMind.
Jurisdictions: Developer: United Kingdom. Operator: United Kingdom.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Each proof is machine-checked in Lean or graded by qualified IMO graders. Problems are unseen by the system before the contest. Grading criteria identical to those for human contestants.
Replication attempts: 1 follow-up entries. Benefit recurred: 1.
Later findings: 21 July 2025. A year later DeepMind's successor system, an advanced version of Gemini Deep Think, scored 35 of 42 at IMO 2025 within the time limit, which the IMO president confirmed as a gold-medal score.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0070-source-0001–0006
Citations: Numbered references below.
Language: Not assessed
Source relationships: Independent, Developer
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: Organizations. Society.
Depth: Not assessed
Reasons: Organizations: A capability evaluation on IMO 2024 problems for the developer and the IMO community. The sources describe grading and scores, not a deployment or an observed organizational change. Society: Proofs of already-solved contest problems graded correct, with no identifiable beneficiary beyond the research community. Scientific American notes around 60 students scored better in 2024.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Entry details

Kind: Not assessed
Related entries: Not assessed
Field: Mathematics competitions. Olympiad mathematics.
Purpose: Solve competition mathematics problems with machine-checkable formal proofs
Dates: Date basis unassessed: 2024 (precision unassessed). Updated: 22 October 2026.
Stage: Not assessed
Version: SHA-256: 4b78415afdaf…

References

Other identifiers: Not assessed
Category mappings: Not assessed
Brief: Not assessed
Revisions: 7 revisions. Latest: 22 October 2026.

Sources

News report Independent author

Bischoff, M. (2024). AI Reaches Silver-Medal Level at This Year's Math Olympiad. Scientific American.
News report Independent author

Williams, R. (2024). Google DeepMind's new AI systems can now solve complex math problems. MIT Technology Review.
News report Independent author

Castelvecchi, D. (2024). DeepMind hits milestone in solving maths problems — AI's next grand challenge. Nature.
News report Independent author

Castelvecchi, D. (2025). DeepMind and OpenAI models solve maths problems at level of top students. Nature.
Press release Developer

Google DeepMind (2024). AI achieves silver-medal standard solving International Mathematical Olympiad problems. Google DeepMind blog.
Press release Developer

Google DeepMind (2025). Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad. Google DeepMind blog.