Post-Cutoff

“We’ve been working on understanding model capabilities beyond what our prior time…”

METR @METR_EvalsX

Archived text

We’ve been working on understanding model capabilities beyond what our prior time horizon tasks could capture.

Today we’re sharing a research update. We find that AI R&D-relevant capabilities are improving rapidly and are getting harder to measure.

Source: x.com/METR_Evals/status/2123864981506798012Archived 2026-11-22 via fxtwitter (unofficial).Counts: 208,635 views, 1,582 likes, 219 reposts, 43 replies (at fetch time)Media: image

Cited in

  1. Benchmarks 143 days after the cutoff

    METR Q4: ~40-hour horizons, judgment lags