Market icon
Tech·AI

AI model scores ≥ 90% on FrontierMath Benchmark before 2027?

Market icon
Tech·AI

AI model scores ≥ 90% on FrontierMath Benchmark before 2027?

Dec 31

Dec 31

14% chance
Polymarket
NEW
14% chance
Polymarket
NEW
This market will resolve to "Yes" if a state-of-the-art (SOTA) AI model achieves a score of 90% or greater on the FrontierMath Exam by December 31, 2026, 11:59 PM ET. Otherwise, the market will resolve to "No". The primary resolution source will be information from EpochAI however a consensus of credible reporting may also be used. Traders assign an 86.5% implied probability to "No" on any AI model achieving ≥90% on the FrontierMath benchmark before 2027, anchored by its November 2024 launch from METR and Epoch AI, which exposed leading large language models' failures—OpenAI's o1 at 1.6%, Anthropic's Claude 3.5 Sonnet at 2.1%, and Google's Gemini 2.0 below 1%. This 200-problem test demands novel mathematical proofs and insights, resisting current scaling laws that propelled prior benchmarks like MATH to ~90%. Absent paradigm-shifting architectures or massive compute leaps, trader consensus reflects cautious extrapolation from demonstrated AI capabilities, though upcoming releases like GPT-5 or Claude 4 could catalyze shifts if they demonstrate frontier math breakthroughs.

This market will resolve to "Yes" if a state-of-the-art (SOTA) AI model achieves a score of 90% or greater on the FrontierMath Exam by December 31, 2026, 11:59 PM ET. Otherwise, the market will resolve to "No".

The primary resolution source will be information from EpochAI however a consensus of credible reporting may also be used.
Volume
$0
End Date
Dec 31, 2026
Market Opened
Nov 12, 2025, 5:15 PM ET

Resolver

0x65070BE91...
This market will resolve to "Yes" if a state-of-the-art (SOTA) AI model achieves a score of 90% or greater on the FrontierMath Exam by December 31, 2026, 11:59 PM ET. Otherwise, the market will resolve to "No". The primary resolution source will be information from EpochAI however a consensus of credible reporting may also be used. Traders assign an 86.5% implied probability to "No" on any AI model achieving ≥90% on the FrontierMath benchmark before 2027, anchored by its November 2024 launch from METR and Epoch AI, which exposed leading large language models' failures—OpenAI's o1 at 1.6%, Anthropic's Claude 3.5 Sonnet at 2.1%, and Google's Gemini 2.0 below 1%. This 200-problem test demands novel mathematical proofs and insights, resisting current scaling laws that propelled prior benchmarks like MATH to ~90%. Absent paradigm-shifting architectures or massive compute leaps, trader consensus reflects cautious extrapolation from demonstrated AI capabilities, though upcoming releases like GPT-5 or Claude 4 could catalyze shifts if they demonstrate frontier math breakthroughs.

This market will resolve to "Yes" if a state-of-the-art (SOTA) AI model achieves a score of 90% or greater on the FrontierMath Exam by December 31, 2026, 11:59 PM ET. Otherwise, the market will resolve to "No".

The primary resolution source will be information from EpochAI however a consensus of credible reporting may also be used.
Volume
$0
End Date
Dec 31, 2026
Market Opened
Nov 12, 2025, 5:15 PM ET

Resolver

0x65070BE91...

Beware of external links.

Frequently Asked Questions

"AI model scores ≥ 90% on FrontierMath Benchmark before 2027?" is a prediction market on Polymarket where traders buy and sell "Yes" or "No" shares based on whether they believe this event will happen. The current crowd-sourced probability is 14% for "Yes." For example, if "Yes" is priced at 14¢, the market collectively assigns a 14% chance that this event will occur. These odds shift continuously as traders react to new developments and information. Shares in the correct outcome are redeemable for $1 each upon market resolution.

"AI model scores ≥ 90% on FrontierMath Benchmark before 2027?" is a newly created market on Polymarket, launched on Nov 12, 2025. As an early market, this is your opportunity to be among the first traders to set the odds and establish the market's initial price signals. You can also bookmark this page to track volume and trading activity as the market gains traction over time.

To trade on "AI model scores ≥ 90% on FrontierMath Benchmark before 2027?," simply choose whether you believe the answer is "Yes" or "No." Each side has a current price that reflects the market's implied probability. Enter your amount and click "Trade." If you buy "Yes" shares and the outcome resolves as "Yes," each share pays out $1. If it resolves as "No," your "Yes" shares pay $0. You can also sell your shares at any time before resolution if you want to lock in a profit or cut a loss.

The current probability for "AI model scores ≥ 90% on FrontierMath Benchmark before 2027?" is 14% for "Yes." This means the Polymarket crowd currently believes there is a 14% chance that this event will occur. These odds update in real-time based on actual trades, providing a continuously updated signal of what the market expects to happen.

The resolution rules for "AI model scores ≥ 90% on FrontierMath Benchmark before 2027?" define exactly what needs to happen for each outcome to be declared a winner — including the official data sources used to determine the result. You can review the complete resolution criteria in the "Rules" section on this page above the comments. We recommend reading the rules carefully before trading, as they specify the precise conditions, edge cases, and sources that govern how this market is settled.