Highest Claude score on Humanity’s Last Exam in 2026?

The central point is that the reported 2026 Claude results are already at the threshold or above it in several recent summaries, with one Claude model listed at 55% and another at 56% on Humanity’s Last Exam. Since the market only needs the highest Claude-family score in 2026 to reach 55% or more, those figures, if reflected on the official leaderboard, are enough to make Yes the natural outcome. This is a materially stronger position than a typical benchmark market where the target has not yet been reached, because the threshold appears to have been crossed already rather than merely approached.

The biggest reason not to assign absolute certainty is that the recent evidence is not perfectly consistent. Different summaries appear to refer to different snapshots, and at least one source gives a lower number for a Claude model or blends in a tools-enabled variant that may not map cleanly to the exact official HLE Accuracy the market will use. In other words, the risk is less about whether Claude can achieve 55% and more about whether the official source and this market’s definition of the metric line up cleanly enough for that score to count. If the leaderboard later shows a lower canonical value, or if the highest score on the official page is reinterpreted as a different variant, the market could still surprise to No despite strong outside evidence.

From a market-pricing perspective, the current Yes price around 87.75% looks somewhat conservative relative to the evidence that the threshold has already been met. That said, prediction markets sometimes leave room for resolution friction, and this event has exactly the sort of wording where ambiguity around metric naming, benchmark variants, or source availability can matter. Because of that, the market should not be treated as a 99% lock, but the base rate and the available public reporting both point heavily toward Yes unless the official leaderboard behaves very differently from the recent summaries.