OpenAIUnited States· released Aug 2025

GPT-5

Name: GPT-5
Author: OpenAI

Best published result on 3 of the 25 tests it has taken.

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

62.0

Ranked #28 (21 to 28) of 36 · 38.0% behind the frontier on average · 25 comparable tests · 4 counting categories

Drop any single test and the Index lands between 59.4 and 65.0, anywhere from #21 to #28. Neighbours inside that range are ties.

8 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, in 25 results.

Strongest in coding (68.6) and agentic (66.7). It holds the best published result on Aider Polyglot, MATH Level 5 and Fiction.LiveBench. It trails the frontier most in reasoning, 45.6% behind on average. Between 35% and 40% behind the best published results, on average. Drop any single test and it would sit anywhere from #21 to #28.

Best published result onAider Polyglot88.0%MATH Level 598.1%Fiction.LiveBench96.9%

Strongest category

Coding

Writing, fixing and shipping real code, judged by tests or users.

68.6

Aider Polyglot88.0% · best
SWE-bench Verified73.6% · 88% of best
SciCode42.9% · 69% of best
WeirdML60.7% · 65% of best
Code Arena (WebDev)1,419 · 20% win rate vs leader

Mean of 5 of the category's 7 comparable tests. Counts toward the Index.

Best single result

Aider Polyglot

Edit code in place to pass the tests of hard exercises across six programming languages.

88.0%

Best published result of 5 models.

Precise, multi-language edits are what a coding assistant does all day inside an editor.

Where it trails

ARC-AGI-2

Infer the hidden rule from a few input and output grid pairs, then apply it to a new grid.

9.9%

89.6% behind the best published result (95.0%) · #42 of 51.

Novel visual puzzles resist memorization, so they measure fluid intelligence rather than recall.

Not measured yet

Covered, but not averaged into the Index:

Knowledgesingle-test category, shown beside the Index · 66.3
Multimodalsingle-test category, shown beside the Index · 71.3
Human preferencesingle-test category, shown beside the Index · 79.4
Long contextsingle-test category, shown beside the Index · 96.9

Every score, with its source

See it at work

What GPT-5 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.

Long context · mean 96.9 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

Fiction.LiveBenchLeaderboard

Answer questions that need the whole of a very long story, at a 120K-token context length.

96.9%

96.9% of the best (100.0%)
#1 of 7

2025-08-07 · medium

Source: Fiction.live

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 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

#28·Index 62.0C

0 : 13

wins · 0 ties · 13 shared

GPT-6 Astra

#1·Index 99.0A+

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

Category means

Reasoning
54.4
99.6
Knowledge
66.3
100.0
Coding
68.6
97.7
Math
58.2
99.8
Agentic
66.7
98.5
Multimodal
71.3
—
Human preference
79.4
—
Long context
96.9
—

Shared tests · biggest gaps first

ARC-AGI-2Reasoning
9.9%
95.0%
−89.6
ProofBenchMath
18.0%
99.0%
−81.0
Code Arena (WebDev)Coding
1,419
1,797
−79.6
FrontierMath Tier 4Math
22.0%
97.6%
−77.5
APEX-AgentsAgentic
18.3%
46.7%
−59.9
CritPtReasoning
12.6%
31.7%
−59.2
FrontierMath (Tiers 1–3)Math
55.4%
93.7%
−40.8
WeirdMLCoding
60.7%
92.9%
−34.7
SimpleQA VerifiedKnowledge
50.1%
75.6%
−33.7
ARC-AGI-1Reasoning
65.7%
98.5%
−33.3
SciCodeCoding
42.9%
56.5%
−21.8
GPQA DiamondReasoning
86.2%
95.8%
−13.6
Mock AIME 2024–2025Math
91.4%
100.0%
−8.6

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

Reasoning54.4 · 6/6
Knowledgebeside66.3 · 1/1
Coding68.6 · 5/7
Math58.2 · 5/5
Agentic66.7 · 5/8
Multimodalbeside71.3 · 1/1
Human preferencebeside79.4 · 1/1
Long contextbeside96.9 · 1/1

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

86.2%

86.4% of the best (95.8%), above chance
38th percentile of 78 · n = 198

2025-08-07 · high

GPT-5

GPT-5, in 25 results.

What GPT-5 was asked to do, and how it did.

GPT-5 against whoever you pick.

Every test, every source

Compare with its neighbours.

The best models, ranked here, working inside Theo.