Observed arrival · 2026-11-04
Frontier Clock maps the lag behind AI’s capability frontier
An interactive analysis that compares AI models’ release dates with the dates they reached the capability level of the leading model.
- For
- Readers tracking AI capability gaps across labs
- Worth noticing
- The page says subset benchmark views are rescaled using its own method, so those values differ from Epoch AI’s official ECI methodology.
Field notes
Models are put on a shared capability scale by fitting scores across overlapping benchmarks; the page describes a sigmoid relationship between model capability and benchmark difficulty. Readers can filter benchmark domains and minimum benchmark counts, choose a running-maximum or interpolated frontier, and inspect 90% intervals produced through repeated benchmark resampling. For benchmark subsets, the page says it linearly rescales values to preserve comparability, noting this is its own method rather than Epoch AI’s official one.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue