Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models
For AI safety researchers, this quantifies the risk that models may evade CoT-based oversight by reasoning internally, providing empirical evidence and projections for a critical monitoring vulnerability.
The paper measures how well frontier AI models reason without chain-of-thought (CoT) across 30,000+ questions, finding that the no-CoT 50% task-completion time horizon has been doubling annually, with GPT-5.5 reaching over 3 minutes and 1,500 reasoning tokens. Projections suggest horizons could exceed 7 minutes by 2028 and 25 minutes by 2030.
Many efforts to ensure frontier AI models are safe rely on monitoring their chain-of-thought (CoT) reasoning. If models become able to perform sufficiently complex reasoning internally, without explicit thinking tokens, this would undermine such oversight. We measure how well frontier models reason without CoT across a suite of over 30,000 questions spanning 43 benchmarks in domains including math, coding, puzzles, causality, theory-of-mind, and strategic reasoning. To compare models against humans, we estimate the $50\%$-task-completion time horizon (TH): the human time required for tasks a model completes with $50\%$ success rate. We complement this with a $50\%$ reasoning token horizon: the minimum number of o3-mini reasoning tokens needed for tasks a model solves with $50\%$ success rate. We find that the no-CoT $50\%$ TH of frontier models has been doubling roughly every year over the past six years, with GPT-5.5's TH reaching over 3 minutes and reasoning token horizon exceeding 1,500 tokens. Our median estimates predict that frontier no-CoT THs could exceed 7 minutes by 2028, and 25 minutes by 2030, though these projections carry substantial uncertainty. We recommend frontier developers track this explicitly.