Highest OpenAI score on Humanity’s Last Exam in 2026?
This event asks whether any OpenAI model will reach a set accuracy on Humanity's Last Exam (HLE) by the end of 2026. Four separate markets use thresholds of 55%, 60%, 65% and 70%. The key rule is that only the 'HLE Accuracy' column on the official leaderboard at agi.safe.ai counts, and the deadline is 11:59 PM Eastern Time on December 31, 2026.
Leading right now: 55%+ 30%, 60%+ 24%, 65%+ 12%, 70%+ 3%.
| Market | Chance of yes | Deadline |
|---|---|---|
| Will the highest score achieved by an OpenAI model on Humanity’s Last Exam in 2026 be 55% or higher? | 29.5% | Jan 1, 2027, 4:59 UTC |
| Will the highest score achieved by an OpenAI model on Humanity’s Last Exam in 2026 be 60% or higher? | 24.5% | Jan 1, 2027, 4:59 UTC |
| Will the highest score achieved by an OpenAI model on Humanity’s Last Exam in 2026 be 65% or higher? | 11.9% | Jan 1, 2027, 4:59 UTC |
| Will the highest score achieved by an OpenAI model on Humanity’s Last Exam in 2026 be 70% or higher? | 3.1% | Jan 1, 2027, 4:59 UTC |
Odds read from Polymarket at 21:54 UTC on Nov 21, 2026. $164K traded so far, $634 in the last 24 hours.
How this market resolves
Yes: Each market resolves YES if any model published by OpenAI reaches at least that market's threshold (55%, 60%, 65% or 70%) in 'HLE Accuracy' on the official HLE leaderboard by December 31, 2026, 11:59 PM ET.
No: A market resolves NO if no OpenAI model reaches its threshold by the deadline. It also resolves NO if the leaderboard is permanently unavailable and no official alternative source exists.
When it really ends
December 31, 2026, 11:59 PM ET (Eastern Time). In December ET is UTC-5, so this is 2027-01-01 04:59 UTC. The endDate field in the data (2027-01-01T04:59:00Z) matches the rules.
What counts
- Any model published by OpenAI, not just its flagship model.
- The value labeled 'HLE Accuracy' on the official leaderboard at agi.safe.ai.
- A clear equivalent metric if the leaderboard renames or restructures its columns.
- A score that meets the threshold exactly ('at least' the stated accuracy).
- Official HLE results published elsewhere, but only if the leaderboard stays unavailable after the timeframe ends or is confirmed permanently gone.
What does not count
- Calibration Error or any other displayed metric.
- Scores from models not published by OpenAI.
- Scores reported only by OpenAI or third parties that are not on the official HLE leaderboard (unless the fallback condition applies).
- Results that appear on the leaderboard only after the deadline, unless they show the model achieved the score by the deadline.
Fine print people miss
- The rules ask whether a score is achieved 'by' the deadline, not whether the score was first set in 2026, even though the title says 'in 2026'. The rules text does not say how a score recorded before 2026 would be treated.
- If the leaderboard goes down during the timeframe, the market stays open until the data is available again.
- If the source is still unavailable after the timeframe, or confirmed permanently gone, other official HLE results may be used.
- If no official alternative exists, the market resolves NO.
- The rules do not say whether tool-using or non-tool-using results are distinguished, or how multiple entries per model are handled. They only point to the 'HLE Accuracy' value.
What moves the price
- A new OpenAI model release with a higher HLE Accuracy on the leaderboard would raise the chance of the higher thresholds resolving YES.
- Leaderboard updates adding OpenAI entries would matter, since only listed results count.
- A long delay in leaderboard updates would lower the chance that a new score is posted before the deadline.
- Changes to the leaderboard's format or hosting could affect resolution; a permanent shutdown with no official alternative points toward NO.
- No new OpenAI model release before the end of 2026 would make the higher thresholds less likely.
Put a bot on this market
A PROWL bot reads the news around the clock, judges each post against these exact rules, and trades inside the limits you set. It starts in paper mode.
Questions people ask
Which score counts for the Humanity's Last Exam OpenAI market?
Only the value labeled 'HLE Accuracy' on the official leaderboard at agi.safe.ai counts. A clear equivalent metric is accepted if the layout or wording changes. Calibration Error and other metrics are ignored.
When does the OpenAI HLE market close?
The deadline is December 31, 2026 at 11:59 PM Eastern Time. That is 04:59 UTC on January 1, 2027.
Does any OpenAI model count, or only the newest one?
Any model published by OpenAI counts. The market asks about the highest score among them.
What happens if the HLE leaderboard goes offline?
The market stays open until the data is available again. If the source is still down after the timeframe ends or is confirmed permanently gone, other official HLE results may be used. If none exist, the market resolves NO.
Are the four thresholds separate markets?
Yes. The 55%, 60%, 65% and 70% markets each resolve on their own against the same leaderboard. A higher score can settle several of them YES at once.
Does OpenAI's own reported score count?
Not by itself. The resolution source is the official leaderboard at agi.safe.ai. Other official HLE results are used only if that leaderboard is unavailable as described in the rules.
More AI markets explained
Written from the market's own rules text on Nov 20, 2026. If the rules on Polymarket differ from this page, the rules on Polymarket govern.