About the Observatory
Review window
The Observatory covers public release materials dated from 2024-01-01 through 2026-09-01, inclusive.
Source boundary
Occurrences require reviewed first-party public materials from the lab responsible for the release. The Observatory records whether those materials name a benchmark and links to the corresponding source.
Score-free purpose
This is a score-free reporting record, not a ranking or comparison of benchmark results.
Reading absence
An absence means only that a benchmark was not found in the reviewed public materials. It does not establish whether a lab evaluated the model privately or reported an evaluation elsewhere.