Compare · ModelsLive · 2 picked · head to head
GPT-5 vs o3
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5 wins on 20/22 benchmarks
GPT-5 wins 20 of 22 shared benchmarks. Leads in coding · reasoning · knowledge.
Category leads
coding·GPT-5reasoning·GPT-5knowledge·GPT-5math·GPT-5
Hype vs Reality
Attention vs performance
GPT-5
#72 by perf·no signal
o3
#67 by perf·no signal
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
22 benchmarks · 2 models
GPT-5o3
Aider polyglot
GPT-5 leads by +6.7
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
GPT-5
88.0
o3
81.3
ARC-AGI
GPT-5 leads by +4.9
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
GPT-5
65.7
o3
60.8
ARC-AGI-2
GPT-5 leads by +3.3
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
GPT-5
9.9
o3
6.5
DeepResearch Bench
GPT-5 leads by +8.5
DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses.
GPT-5
55.1
o3
46.6
Fiction.LiveBench
GPT-5 leads by +8.3
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
GPT-5
97.2
o3
88.9
FrontierMath-2025-02-28-Private
GPT-5 leads by +13.7
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
GPT-5
32.4
o3
18.7
FrontierMath-Tier-4-2025-07-01-Private
GPT-5 leads by +10.4
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
GPT-5
12.5
o3
2.1
GeoBench
GPT-5 leads by +7.0
GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding.
GPT-5
81.0
o3
74.0
GPQA diamond
GPT-5 leads by +5.8
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GPT-5
81.6
o3
75.8
GSO-Bench
o3 leads by +1.9
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues.
GPT-5
6.9
o3
8.8
HLE
GPT-5 leads by +5.3
HLE (Humanity's Last Exam) · crowdsourced expert-level questions designed to be among the hardest possible challenges for AI systems across all domains.
GPT-5
21.6
o3
16.3
Lech Mazur Writing
GPT-5 leads by +3.3
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
GPT-5
87.2
o3
83.9
MATH level 5
GPT-5 leads by +0.4
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
GPT-5
98.1
o3
97.8
OTIS Mock AIME 2024-2025
GPT-5 leads by +7.5
OTIS Mock AIME 2024–2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GPT-5
91.4
o3
83.9
Professional Reasoning · Finance
GPT-5 leads by +3.6
GPT-5
51.3
o3
47.7
Professional Reasoning · Legal
GPT-5 leads by +0.4
GPT-5
49.0
o3
48.6
SimpleBench
GPT-5 leads by +4.3
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
GPT-5
48.0
o3
43.7
SimpleQA Verified
o3 leads by +2.4
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
GPT-5
50.6
o3
53.0
SWE-Bench verified
GPT-5 leads by +11.2
GPT-5
73.5
o3
62.3
SWE-Bench Verified (Bash Only)
GPT-5 leads by +6.6
SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks.
GPT-5
65.0
o3
58.4
VPCT
GPT-5 leads by +21.0
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
GPT-5
49.0
o3
28.0
WeirdML
GPT-5 leads by +8.3
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
GPT-5
60.7
o3
52.4
Full benchmark table
| Benchmark | GPT-5 | o3 |
|---|---|---|
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 88.0 | 81.3 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 65.7 | 60.8 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 9.9 | 6.5 |
DeepResearch Bench DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses. | 55.1 | 46.6 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 97.2 | 88.9 |
FrontierMath-2025-02-28-Private FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning. | 32.4 | 18.7 |
FrontierMath-Tier-4-2025-07-01-Private FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning. | 12.5 | 2.1 |
GeoBench GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding. | 81.0 | 74.0 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 81.6 | 75.8 |
GSO-Bench GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues. | 6.9 | 8.8 |
HLE HLE (Humanity's Last Exam) · crowdsourced expert-level questions designed to be among the hardest possible challenges for AI systems across all domains. | 21.6 | 16.3 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 87.2 | 83.9 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 98.1 | 97.8 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024–2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 91.4 | 83.9 |
Professional Reasoning · Finance | 51.3 | 47.7 |
Professional Reasoning · Legal | 49.0 | 48.6 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 48.0 | 43.7 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 50.6 | 53.0 |
SWE-Bench verified | 73.5 | 62.3 |
SWE-Bench Verified (Bash Only) SWE-Bench Verified (Bash Only) · a curated subset of SWE-bench where models fix real Python repository bugs using only bash commands, no agent frameworks. | 65.0 | 58.4 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | 49.0 | 28.0 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 60.7 | 52.4 |
Pricing · per 1M tokens · projected $/mo at 10M tokens