B.O.B. Leaderboard

OpenAIreleased 5 Mar 2026

GPT-5.4 Pro

#16 of 9513 of 46 benchmarksHolds up

70.4

B.O.B.

Category profile

  • Real use (hard to game)2×74
  • Reasoning & knowledge72
  • Math73
  • Coding45
  • Agents & real work—
  • Independent indexes85

Reality check

Lab tests

63

Real use (2 tests)

74

Does even better in real use than on lab tests. It does 10 points better on tests that are hard to train for than on the usual lab tests.

What counts as real use

Price per 1M tokens (OpenRouter)

$30.00 in · $180.00 out

Cost per task on the AA index

—

Output speed (Artificial Analysis)

—

§Receipts

Every benchmark result

“Beats” = the share of tracked models this one beat on that test. The raw score is its best reported setting.

BenchmarkRawBeatsSource
Real use (hard to game) · Real use 74
SimpleBenchas gpt-5.4-pro-2026-03-05_unknown74.1%
76%of 30
Epoch AI Benchmarking Hub
SkateBenchas gpt-5.4-pro-thinking79.0%
71%of 8
Theo (t3.gg, YouTube)
Reasoning & knowledge · Reasoning 72
GPQA Diamondas gpt-5.4-pro-2026-03-05_xhigh94.6%
91%of 70
Epoch AI Benchmarking Hub
CritPt (physics research)as gpt-5.4-pro-2026-03-05_xhigh30.0%
90%of 89
Epoch AI Benchmarking Hub
Chess Puzzlesas gpt-5.4-pro-2026-03-05_xhigh59.0%
90%of 73
Epoch AI Benchmarking Hub
ARC-AGI-2as gpt-5.4-pro-2026-03-05_xhigh83.3%
67%of 44
Epoch AI Benchmarking Hub
Humanity's Last Examas gpt-5.4-pro-2026-03-05_unknown44.3%
58%of 13
Epoch AI Benchmarking Hub
EnigmaEvalas gpt-5.4-pro-2026-03-05_unknown23.8%
56%of 10
Epoch AI Benchmarking Hub
SimpleQA Verifiedas gpt-5.4-pro-2026-03-05_xhigh46.3%
48%of 57
Epoch AI Benchmarking Hub
Math · Math 73
FrontierMath Tiers 1-3as gpt-5.4-pro-2026-03-05_xhigh82.5%
78%of 60
Epoch AI Benchmarking Hub
FrontierMath Tier 4as gpt-5.4-pro-2026-03-05_xhigh58.5%
69%of 52
Epoch AI Benchmarking Hub
Coding · Coding 45
WeirdMLas gpt-5.4-pro-2026-03-05_none57.4%
45%of 52
Epoch AI Benchmarking Hub
Independent indexes · Indexes 85
Epoch Capabilities Indexas GPT-5.4 Pro159
85%of 81
Epoch AI