METR evaluated an early version of Claude Mythos

[https://metr.org/time-horizons/](https://metr.org/time-horizons/) "We evaluated an early version of Claude Mythos Preview for risk assessment during a limited window in March 2026. We estimated a 50%-time-horizon of at least 16hrs (95% CI 8.5hrs to 55hrs) on our task suite, at the upper end of what we can measure without new tasks. [](https://x.com/METR_Evals/status/2052896621760004602/photo/1) Of the 228 tasks in our suite, only 5 are estimated as 16+ hours long, making measurements at this range unstable and less meaningful than at ranges with better task coverage. Thus, we are not highlighting exact estimates for models above 16 hours measured with our current suite. [](https://x.com/METR_Evals/status/2052896623852929510/photo/1) We believe that this task suite could still distinguish a much more capable model from current publicly-known state-of-the-art models. But we do not consider measurements at this range to be robust enough for precise quantitative comparisons or extrapolations. In principle the time-horizon methodology allows us to measure higher capability models by adding longer tasks, and we’re working on updated methods. But these are still in development; for now, we advise caution in interpreting recent time-horizon numbers."