The 'Humanity’s Last Exam' is designed to test the limits of AI reasoning across a broad range of subjects, with a 50% threshold seen as a significant bar for general intelligence. Polymarket, the only platform tracking this event with $27,000 in volume, shows a 33% probability that OpenAI’s GPT will achieve this score by the resolution date of June 30, 2026. This implies a 67% chance the model will not reach the mark, reflecting cautious expectations among traders about near-term AI capabilities.
No data is available from Kalshi or Manifold, limiting cross-platform comparison. The relatively low volume suggests moderate interest, and the absence of recent news headlines means no external triggers have shifted the odds. The market’s assessment aligns with broader skepticism in the AI community about achieving human-level reasoning on such a comprehensive test within the next two years.
Traders appear to weigh the rapid pace of AI development against the exam’s difficulty, which includes complex problem-solving and nuanced understanding. The 33% probability indicates a non-trivial chance of success, but the majority expectation is that GPT will not cross the 50% threshold, underscoring the challenge of surpassing this benchmark.