Highest score on Humanity’s Last Exam in 2026?
The market is strongly tilted toward Yes because a 2026 result already appears to have crossed 60% on Humanity’s Last Exam. The main remaining uncertainty is benchmark framing and source matching, not whether the threshold itself has been surpassed.
Analysis
The central fact is that the 60% threshold appears to have already been beaten in 2026, with Claude Opus 5.5 reported at 67.7% with tools and 61.4% on a text-only or closely related reporting variant. Because the market resolves based on the highest score achieved during 2026, a qualifying score already appearing in the record makes Yes the dominant outcome unless there is a later dispute that the reported metric does not count as the official HLE Accuracy measure.
The strongest argument for caution is that benchmark naming and presentation can be messy. Some figures are reported with tools enabled, some without tools, and some on variants or derivative leaderboards like HLE-Diamond, so a strict reading of the market could exclude scores that are not on the official leaderboard or are not directly comparable. Even so, the recent context suggests the official and semi-official reporting ecosystem is converging around scores above 60%, which makes it unlikely that the year ends below the cutoff unless the cited numbers are later invalidated or reclassified.
Market pricing at 60% Yes still looks slightly conservative relative to the evidence. The current price likely reflects residual uncertainty about whether the official resolution source will mirror the public claims cleanly, but the existence of multiple 2026 scores above 60% means the distribution of outcomes is heavily skewed toward Yes. To overturn Yes, one would need a strong reason to believe the reported benchmark results do not count for resolution or that all above-60% claims are ultimately disqualified.
Historically, frontier benchmark gains have tended to compound rather than reverse, especially when multiple labs are competing on the same task. Once a model family posts a result above a well-publicized threshold, later improvements, reruns, or alternate evaluation settings usually preserve that level rather than erasing it. That makes the practical probability of the year-end maximum staying at or above 60% very high, with the remaining uncertainty concentrated in definitional edge cases rather than performance trends.
Arguments
For
- Arguments for Yes: A reported 2026 result already exceeds 60%, which is enough if the official leaderboard accepts that score.
- Arguments for Yes: The benchmark trend suggests continued progress, so it is more likely the year-end maximum remains above the threshold than drops below it.
Against
- Arguments against Yes: The public reporting mixes tool-augmented, text-only, and variant-specific scores, which could complicate strict resolution.
- Arguments against Yes: If the official leaderboard does not validate the reported number, the market may have to fall back to a lower officially recognized score.
Key drivers
- A 2026-reported HLE score above 60% already exists, which directly satisfies the market condition if it is accepted as official.
- Multiple independent writeups and leaderboard-style reports point to consistent benchmark results above the threshold.
- The competitive frontier in reasoning benchmarks is advancing quickly, making a year-end maximum below 60% increasingly implausible.
Risk factors
- The official resolution source may distinguish between benchmark variants, tools-enabled runs, and the exact HLE Accuracy metric.
- If the above-60% result is later deemed non-comparable or not officially recognized, the market could revert to a lower valid maximum.
Scenarios
Best case
The official Humanity’s Last Exam leaderboard clearly records a 2026 score above 60% and possibly above the mid-60s, removing any ambiguity and making Yes undisputed.
Most likely
The reported above-60% score is accepted in substance, and the year closes with the highest 2026 Humanity’s Last Exam score comfortably at or above 60%.
Worst case
The above-60% claims are treated as non-comparable or not official, and the highest accepted official score for 2026 ends up below 60%, forcing a No resolution.
More from this day
- economyPolymarket5d
No change flips Hike 25 bps for October Fed by...?
AI22%MKT7%Edge+15Hidden GemThe market currently leans toward another 25 bps hike, so the near-term challenge is not just a brief dip in hike odds but four consecutive hourly prints with no change above hike. That makes the Yes outcome possible if the September 30 PCE report softens expectations, but it still looks meaningfully less likely than No.
- cryptoPolymarket3mo
Extended FDV above ___ one day after launch?
AI38%MKT24%Edge+14UndervaluedI lean slightly toward Yes, but not strongly. The main reason is that if Extended does launch a token, the market is already signaling a valuation regime where a $500M FDV is not especially demanding, though the biggest uncertainty is whether a launch happens at all and whether the initial trading price holds up one day later.