Claim · reasoning

GPT-5.4 (2026-03-05)on GPQA Diamond (Epoch's own run)

74.8 to 93.3 the range across every scoreable measurement we hold

18.6 points between the highest and the lowest

88.9 the number the board uses: the lower median, a real row with its own provenance

5 declared configurations, from 1 runner

All 5 scoreable measurements here come from ONE runner, so this is not a reconciliation and is not presented as one. What it shows is how far a single runner's number moves when the model is run differently.

Every measurement

5 rows. Best first. the board takes the LOWER MEDIAN of the scoreable measurements, which is a real row with its own interval and provenance rather than an interpolated average. a vendor self-report is shown beside the independent numbers and is never one of them; it cannot move a rank.

Score	Provenance	Configuration	Who ran it	Where to check it
93.3 / 100 as published: 93.3%	INDEPENDENT counts toward the ranking	gpt-5.4-2026-03-05 effort xhigh	Epoch AI 2026-03-06T02:57:02.864Z CC-BY-4.0	NO_URL2026-08-30 benchmark leaderboard the leaderboard for this benchmark, not a link to this row
89.9 / 100 as published: 89.9%	INDEPENDENT counts toward the ranking	gpt-5.4-2026-03-05 effort high	Epoch AI 2026-07-15T03:14:50.000Z CC-BY-4.0	NO_URL2026-08-30 benchmark leaderboard the leaderboard for this benchmark, not a link to this row
88.9 / 100 as published: 88.9%	INDEPENDENT counts toward the ranking	gpt-5.4-2026-03-05 effort medium	Epoch AI 2026-07-15T03:14:58.000Z CC-BY-4.0	NO_URL2026-08-30 benchmark leaderboard the leaderboard for this benchmark, not a link to this row
84.8 / 100 as published: 84.8%	INDEPENDENT counts toward the ranking	gpt-5.4-2026-03-05 effort low	Epoch AI 2026-07-15T03:14:49.000Z CC-BY-4.0	NO_URL2026-08-30 benchmark leaderboard the leaderboard for this benchmark, not a link to this row
74.8 / 100 as published: 74.7%	INDEPENDENT counts toward the ranking	gpt-5.4-2026-03-05 effort none	Epoch AI 2026-07-15T03:14:58.000Z CC-BY-4.0	NO_URL2026-08-30 benchmark leaderboard the leaderboard for this benchmark, not a link to this row

← The Forge Rankings

Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e, as of 2026-08-30. where only one party has measured a model on a benchmark, this page says so rather than presenting a single number as a reconciliation.