This prediction was committed to a SHA-256 hash at call time. Anyone can verify the call was not changed after the fact.
Will an Anthropic Claude model score at least 55% on Humanity’s Last Exam? | No | 2026-06-30
65de9514
Verifying…
echo -n "Will an Anthropic Claude model score at least 55% on Humanity’s Last Exam? | No | 2026-06-30" | shasum -a 256 | cut -c1-8Anthropic's public Claude models (Opus 4.7 etc.) sit at ~36-40% on the official Scale HLE leaderboard with no leaks, announcements, or procedural signals of a new release capable of a 15+ point jump in the next 5 weeks. Mythos Preview's high internal scores do not appear on the resolution leaderboard and are explicitly non-public. Decision-makers' recent behavior (April 2026 releases only) and base rates for frontier benchmark leaps contradict the Yes thesis; smart-money whales correctly price this near-certain No.
Official Scale leaderboard (the sole resolution source) shows top Claude models at 36.2% (Opus 4.7) and 34.4% (Opus 4.6 variants) as of latest data, with no model near 55%. Market rules require the score to appear on that exact leaderboard by 11:59 PM ET June 30, 2026; no resolver updates override this. With ~5 weeks remaining and no confirmed releases or updates capable of a 15+ point jump, the probability of Yes is well below 10%, satisfying the >90% certainty bar for the recommended No side.
Market microstructure supports the No side: Yes priced at 6.1¢ with only $4.4K total volume since April open, stable 1w/1m price action near that level, and smart-money whales (including high-PnL accounts staking hundreds on No) aligned at confidence 1.0 with no opposing flow. Thin liquidity and low weekly turnover are consistent with a high-certainty consensus hold rather than an unarbed edge, and no recent price drift or sibling-bin contradiction appears.
+2 more open picks · full 3-judge reasoning · Telegram premium channel.
Subscribe Now