A physics benchmark goes from 9% to 30% in four months

Claim

GPT-5.4 Pro scored 30% on the CritPt physics benchmark, a ten-point gain on a test where the best score was 9% four months earlier.

Reported metric: 9% → 30%

Reporting date: 2026-03-08 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to March 8, 2026 (theinnermostloop.substack.com)