Highest Claude score on Humanity’s Last Exam in 2026?: how this market works

What you need to know

This market asks how high a score Claude, the AI made by Anthropic, will reach on a famously difficult test called Humanity's Last Exam. That test is a collection of extremely hard questions across math, science, and other fields, designed so that even top AI systems struggle with it. A score of 55% means Claude answers more than half correctly, 60% means three out of five, and 65% means nearly two out of three. The higher the threshold, the harder the target to hit. Each of the three versions of this market resolves Yes if any Claude model hits that specific accuracy threshold on the official Humanity's Last Exam leaderboard before December 31, 2026. The score that counts is the one labeled 'HLE Accuracy' on the leaderboard at agi.safe.ai. Only that metric matters, not related scores like calibration error. If the leaderboard goes permanently offline and no official alternative exists, all versions resolve No by default. The news provided does not contain anything relevant to this market. There are no recent headlines about Claude, Anthropic, or Humanity's Last Exam in the list given. To follow this question, the most useful things to watch for would be Anthropic releasing new Claude models and any updated scores appearing on the official HLE leaderboard. AI benchmark progress is genuinely hard to forecast because it depends on when Anthropic releases new models and how much those models improve. The market already prices the 55% bar at 90%, suggesting that one is considered fairly achievable. But the 65% bar sits at only 49%, reflecting real disagreement about whether that leap happens within 2026. The pace of AI improvement has surprised experts in both directions, so the timeline and magnitude of gains remain genuinely open questions.

The odds right now

• 55%+90%
• 60%+34%
• 65%+32%
• 70%+20%
• 75%+5%

Price history

How this resolves

Resolves December 31, 2026

This market will resolve to "Yes" if any Anthropic Claude model achieves at least the specified accuracy on Humanity’s Last Exam by December 31, 2026, 11:59 PM ET. Otherwise, this market will resolve to "No". Read the full resolution rules on the live market page.

Related

Same markets. A fraction of the fee.

These apps all route to the same exchange order book. The difference is what each one adds on top of the exchange's own fee.

Published rates, checked July 2026. MetaMask charges a flat 4 percent per prediction trade. Jupiter adds a fee equal to the exchange's own taker fee at fill time, roughly 2 to 4 percent at typical odds. Where a market carries an exchange settlement fee, it applies everywhere, whichever app you use. Paridesk adds nothing on maker orders.

Trade this market on Paridesk: non-custodial, 0.5% fee.