Claude Hit 30% on Humanity’s Last Exam by June 30 | Lines.com

Claude crossed the 30 percent threshold on Humanity’s Last Exam before the June 30, 2026 deadline, resolving this Polymarket contract at YES. Anthropic’s model cleared the benchmark that frontier AI had treated as a near-impossible ceiling just eighteen months earlier. The result confirmed what a $407,179 market had priced as a mathematical certainty for weeks.

The implied probability on this market sat at 100 percent from the opening trade. Traders never wavered. The final price at close matched the opening price: 1.00. That kind of unanimous conviction is rare on a science and technology benchmark, and it proved correct. The market did not predict this outcome so much as it ratified it.

Claude Clears 30 Percent on the Hardest AI Benchmark

Humanity’s Last Exam launched in early 2025 as a collaboration between the Center for AI Safety and Scale AI. The benchmark contains roughly 3,000 expert-level questions spanning mathematics, science, law, and humanities. When frontier models first faced HLE, top scores clustered below 10 percent. Claude 3.5 Sonnet scored approximately 4.3 percent at launch. The 30 percent bar looked distant.

By mid-2026, Anthropic’s Claude had closed that gap decisively. The model’s score crossed 30 percent before the June 30 deadline, triggering resolution. The jump reflects architectural improvements, extended reasoning capabilities, and a scaling trajectory that outpaced early HLE projections by a wide margin.

The final probability at close held at 1.00. No late movement, no hedging, no correction. Traders who entered this market early locked in a confirmed result and never had reason to exit at a discount.

Sponsored Partner

How the Market Performed

What the Pricing Said About Trader Conviction

The implied probability never dipped below 100 percent across the observable price history. That means traders treated a Claude score of 30 percent or higher on HLE as a resolved fact, not a probabilistic forecast. Markets that price at 1.00 and hold there typically reflect one of two conditions: the outcome was already confirmed when trading began, or the directional case was so overwhelming that no credible counterargument existed.

Total volume reached $407,179 across the market’s life. The $45,325 liquidity pool and $16,549 in 24-hour volume near resolution suggest active participation even after the outcome became clear. Traders continued cycling through the contract, consistent with a market that had already found its floor and ceiling at the same level.

What Claude’s HLE Score Means for AI Benchmarking

A 30 percent score on Humanity’s Last Exam is not a rounding error. HLE was designed explicitly to resist saturation, drawing on questions that required genuine expert-level reasoning rather than pattern matching against training data. Claude clearing 30 percent signals that frontier AI has moved into a performance tier that HLE’s designers anticipated would take longer to reach.

For the prediction market structure, this contract illustrates both the value and the limitation of binary AI benchmarks. The YES outcome was correct, but the 30 percent threshold captured only the floor of potential outcomes. The market offered four higher tiers: 35 percent, 45 percent, 50 percent, and 55 percent. Where Claude actually landed within or above that range carries more analytical weight than the binary resolution alone.

Forward signals from this resolution:

• Anthropic’s Claude series now benchmarks above the 30 percent HLE threshold, setting a new floor for competitive comparisons against OpenAI and Google DeepMind models.
• HLE will likely require recalibration as a frontier benchmark if multiple models approach or exceed 30 percent within the same measurement window.
• Related markets tracking AI capability milestones, including OpenAI IPO timing and AI bubble risk, now operate against a backdrop of measurably accelerating model performance.
• Future HLE-linked prediction markets should anchor to higher thresholds, as the 30 percent bar no longer represents genuine uncertainty for leading frontier models.

LINES RESOLUTION VERDICT

CONFIRMED YES: CORRECTLY PRICED

Claude hit 30 percent on Humanity’s Last Exam by June 30, 2026, and the market never pretended otherwise — 100 percent pricing from open to close is not a forecast, it is a consensus, and this one was right.

What the market showed: The implied probability held at 100 percent throughout. The final price at close matched: 1.00. Traders priced this as a certainty and the outcome confirmed it. The math doesn’t lie — when a market never moves off 1.00, it is either reckless or informed, and here it was informed.