OpenAIUnited States· released Mar 2026

GPT-5.4

Name: GPT-5.4
Author: OpenAI

Ranked #10, 22.7% behind the frontier on average.

OpenAI, United States. One of 26 models from this lab on the Index.

Third-party publishedHigh confidence (12+ comparable tests, all 4 Index categories)

Snapshot September 8, 2026

OpenCharts Index

77.3

Ranked #10 (10 to 11) of 36 · 22.7% behind the frontier on average · 21 comparable tests · 4 counting categories

Drop any single test and the Index lands between 75.2 and 79.1, anywhere from #10 to #11. Neighbours inside that range are ties.

7 of 8 hardest tests · share of the frontier

Distance to today's frontier on the hardest tests. Not a measure of intelligence.

What it's known for

GPT-5.4, in 21 results.

Strongest in reasoning (84.0) and agentic (80.1). It trails the frontier most in math, 28.0% behind on average. No comparable result yet in long context. Between 20% and 25% behind the best published results, on average. Drop any single test and it would sit anywhere from #10 to #11.

Strongest category

Reasoning

Hard, novel problems: graduate science, abstract puzzles, expert exams.

84.0

GPQA Diamond93.3% · 97% of best
ARC-AGI-193.7% · 95% of best
Humanity's Last Exam36.2% · 78% of best
ARC-AGI-274.0% · 78% of best
CritPt23.4% · 73% of best

Mean of 5 of the category's 6 comparable tests. Counts toward the Index.

Best single result

Mock AIME 2024–2025

Solve competition mathematics problems that have a single exact integer answer.

97.8%

97.8% of the best (100.0%) · #12 of 78.

Olympiad-style problems need long, careful chains of reasoning, and there is no partial credit.

Where it trails

Code Arena (WebDev)

Build a web app from the same prompt as a rival model; real users vote blind on the result.

1,463

74.5 points short of parity with the board leader (1,797) · #41 of 86.

People judging finished apps side by side is the most honest measure of front-end quality.

Not measured yet

No comparable result yet in this category. A blank is a blank, never a zero, and it does not lower the Index.

Long context

Covered, but not averaged into the Index:

Knowledgesingle-test category, shown beside the Index · 59.7
Multimodalsingle-test category, shown beside the Index · 90.7
Human preferencesingle-test category, shown beside the Index · 91.2

Every score, with its source

See it at work

What GPT-5.4 was asked to do, and how it did.

Every test it has taken, by category, with the kind of task it faced, who produced the number and where the result landed against the best published one. The examples are original and illustrative, never items from the datasets themselves.

Human preference · mean 91.2 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

Text ArenaLeaderboard

Answer the same prompt as a rival model; real users vote blind on which answer they prefer.

1,477

46% expected win rate against the board leader (1,507)
#21 of 92

high · board 2026-09-02

Source: Arena

Every result is read against the best published one on its test: percent scores as a share of the best above chance, Arena ratings as the expected win rate against the board leader, open-ended values on a log scale. Sitting a harder exam never lowers a model. Every number links to the publisher that produced it.

Head to head

GPT-5.4 against whoever you pick.

Choose a rival. Every counted test both have taken appears side by side as a share of the best published result on that test, with the category means above. A win is a higher share; a gap under 2 points is a tie, because one item on a 45-problem exam is 2.2 points and publishers' own standard errors are of that order.

GPT-5.4

#10·Index 77.3B

0 : 12

wins · 1 tie · 13 shared

GPT-6 Astra

#1·Index 99.0A+

GPT-6 Astra comes out ahead on 12 of the 13 tests both have taken.

Category means

Reasoning
84.0
99.6
Knowledge
59.7
100.0
Coding
73.1
97.7
Math
72.0
99.8
Agentic
80.1
98.5
Multimodal
90.7
—
Human preference
91.2
—

Shared tests · biggest gaps first

Code Arena (WebDev)Coding
1,463
1,797
−74.5
FrontierMath Tier 4Math
49.0%
97.6%
−49.8
ProofBenchMath
56.0%
99.0%
−43.0
SimpleQA VerifiedKnowledge
45.1%
75.6%
−40.3
CritPtReasoning
23.4%
31.7%
−25.7
APEX-AgentsAgentic
36.0%
46.7%
−22.6
ARC-AGI-2Reasoning
74.0%
95.0%
−22.2
WeirdMLCoding
77.7%
92.9%
−16.4
FrontierMath (Tiers 1–3)Math
78.6%
93.7%
−16.1
ARC-AGI-1Reasoning
93.7%
98.5%
−4.9
GPQA DiamondReasoning
93.3%
95.8%
−3.5
Mock AIME 2024–2025Math
97.8%
100.0%
−2.2
SciCodeCoding
56.6%
56.5%
tie

Bars are distances to the best published result on each test (percent scores above chance, Arena ratings as win rate against the leader, open-ended values on a log scale). Raw values as the publishers report them.

Open GPT-6 Astra

Scores

Every test, every source

The best published run per benchmark, the raw value as the publisher reports it, who produced the number, how close it comes to the best published result on that test, and where the model sits among every model scored on it. Percent scores are read above chance, Arena ratings as a win rate against the board leader, open-ended values on a log scale. A test with too few models to compare against is shown but not counted.

Category means

Reasoning84.0 · 5/6
Knowledgebeside59.7 · 1/1
Coding73.1 · 4/7
Math72.0 · 4/5
Agentic80.1 · 5/8
Multimodalbeside90.7 · 1/1
Human preferencebeside91.2 · 1/1
Long context—

Mean · tests taken of the category's comparable tests. A faded category either rests on fewer than half of them or is a single-test category (marked “beside”), which is shown next to the Index and never averaged into it.

GPQA DiamondEpoch-runReasoning

93.3%

96.5% of the best (95.8%), above chance
87th percentile of 78 · n = 198

2026-03-05 · xhigh

GPT-5.4

GPT-5.4, in 21 results.

What GPT-5.4 was asked to do, and how it did.

GPT-5.4 against whoever you pick.

Every test, every source

Compare with its neighbours.

The best models, ranked here, working inside Theo.