On Polymarket, the only platform currently tracking this event, traders have priced in a low probability—just 11%—that OpenAI’s forthcoming GPT system will achieve a passing grade on Humanity’s Last Exam, a rigorous test of general intelligence and problem-solving. The market, with $25,000 in volume, reflects a consensus that the model will likely fall short, though the modest liquidity suggests limited conviction. No other prediction markets, such as Kalshi or Manifold, have listed contracts on this specific outcome, leaving Polymarket as the sole gauge of expectations.
The exam, designed by AI researchers to probe the boundaries of machine cognition, covers topics from mathematics to ethics, and a 50% threshold is seen as a significant milestone. The 11% probability implies that traders view the task as extremely challenging for current architectures, even as OpenAI continues to push the frontier with each iteration. Without recent news headlines to explain the odds, the market appears to reflect a baseline skepticism about near-term AI breakthroughs, rather than a reaction to any specific event.
This low probability aligns with broader trends in AI forecasting, where markets have historically underestimated rapid progress. However, the absence of cross-platform data limits the robustness of the signal, and the relatively small volume means the odds could shift quickly with new developments, such as a leaked benchmark result or a major technical paper from OpenAI.