In what year will AI achieve a score of 95% or higher on the USAMO benchmark?
3
225Ṁ149
2033
November 1, 2028
8%
2025
8%
2026
20%
2027
22%
2028
12%
2029
12%
2030
8%
2031
8%
2032

Background

  1. What is USAMO? The USA Mathematical Olympiad is a two-day proof contest consisting of six problems worth 7 points each (42 points total). It is widely regarded as the hardest high-school math exam in the United States

  2. Why it matters: Unlike short-answer math benchmarks (e.g. GSM8K), USAMO requires multi-page proofs—demanding creativity, rigor, and long-horizon reasoning similar to the International Math Olympiad

  3. Benchmarking platform: The MathArena project publishes uncontaminated, post-release leaderboards for new math competitions, grading each AI solution multiple times and reporting pass@1 style accuracies

State of Play:

  1. DeepSeek-R1-0528: 30.1%

  2. Gemini 2.5 Pro: 24.4%

  3. Human: ≳ 90%

Why this milestone matters

  • Proof-level mastery: 95 % implies solving at least five of six Olympiad proofs matching elite human contestants

  • Economic & scientific spill-overs: Breakthroughs in formal proof and symbolic reasoning could accelerate research automation in STEM.

Resolution Criteria

This market resolves to the calendar year in which ALL of the following occur:

  1. Score ≥ 95 % on any official USAMO held after the model’s public release.

  2. Verification — the result is confirmed by either

    1. a peer-reviewed paper on arXiv, or

    2. an official MathArena leaderboard entry or an equivalently rigorous public board.

  3. Autonomy — unlimited compute/tools are fine; no hidden human guidance.

Fine print

  • If no qualifying run is verified by Jan 1, 2033, the market resolves to “Not Applicable.”

Get
Ṁ1,000
to start trading!
© Manifold Markets, Inc.TermsPrivacy