⌕Is ___ SOTA?
benchmarks / metr-time-horizons

METR Time Horizons

time horizon (minutes). Status display_only. Contamination risk low. Maintainer METR. Source.

Display only.This benchmark cannot decide a crown under methodology v1.0.
ModelScoreDateKind
Claude Opus 4.6718.806832026-02-05observation
GPT-5.2352.2493022025-12-11observation
GPT-5.3 Codex349.5307322026-02-05observation
Claude Opus 4.5292.9945942025-11-24observation
Gemini 3 Pro Preview224.3258842025-11-18observation
GPT-5203.0125772025-08-07observation
O3119.7326342025-04-16observation
Claude Opus 4.1100.4720042025-08-05observation
Claude Opus 4100.3661232025-05-22observation
Claude 3.7 Sonnet60.3889372025-02-24observation
GPT-4O6.9911952024-05-13observation
GPT-4 03143.9874282023-03-14observation
GPT-4 Turbo3.7327872024-04-09observation