Compare · ModelsLive · 2 picked · head to head
Claude Opus 4 vs GPT-5
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5 wins on 15/18 benchmarks
GPT-5 wins 15 of 18 shared benchmarks. Leads in coding · reasoning · knowledge.
Category leads
coding·GPT-5reasoning·GPT-5knowledge·GPT-5math·GPT-5
Hype vs Reality
Attention vs performance
Claude Opus 4
#131 by perf·no signal
GPT-5
#72 by perf·no signal
Best value
GPT-5
10.4x better value than Claude Opus 4
Claude Opus 4
0.9 pts/$
$45.00/M
GPT-5
9.7 pts/$
$5.63/M
Vendor risk
Who is behind the model
Anthropic
$380.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
18 benchmarks · 2 models
Claude Opus 4GPT-5
Aider polyglot
GPT-5 leads by +16.0
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
Claude Opus 4
72.0
GPT-5
88.0
ARC-AGI
GPT-5 leads by +30.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Opus 4
35.7
GPT-5
65.7
ARC-AGI-2
GPT-5 leads by +1.3
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Claude Opus 4
8.6
GPT-5
9.9
DeepResearch Bench
GPT-5 leads by +6.1
DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses.
Claude Opus 4
49.0
GPT-5
55.1
Fiction.LiveBench
GPT-5 leads by +36.1
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
Claude Opus 4
61.1
GPT-5
97.2
FrontierMath-2025-02-28-Private
GPT-5 leads by +27.9
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Claude Opus 4
4.5
GPT-5
32.4
FrontierMath-Tier-4-2025-07-01-Private
GPT-5 leads by +8.3
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
Claude Opus 4
4.2
GPT-5
12.5
GeoBench
GPT-5 leads by +32.0
GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding.
Claude Opus 4
49.0
GPT-5
81.0
GPQA diamond
GPT-5 leads by +13.2
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Opus 4
68.3
GPT-5
81.6
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.
Claude Opus 4
6.9
GPT-5
6.9
HLE
GPT-5 leads by +15.3
HLE (Humanity's Last Exam) · crowdsourced expert-level questions designed to be among the hardest possible challenges for AI systems across all domains.
Claude Opus 4
6.2
GPT-5
21.6
MATH level 5
GPT-5 leads by +13.1
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Claude Opus 4
85.0
GPT-5
98.1
OTIS Mock AIME 2024-2025
GPT-5 leads by +27.0
OTIS Mock AIME 2024–2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Opus 4
64.4
GPT-5
91.4
SimpleBench
Claude Opus 4 leads by +2.5
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Claude Opus 4
50.6
GPT-5
48.0
SWE-Bench verified
GPT-5 leads by +2.9
Claude Opus 4
70.7
GPT-5
73.5
SWE-Bench Verified (Bash Only)
Claude Opus 4 leads by +2.6
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.
Claude Opus 4
67.6
GPT-5
65.0
VPCT
GPT-5 leads by +42.0
VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations.
Claude Opus 4
7.0
GPT-5
49.0
WeirdML
GPT-5 leads by +17.3
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude Opus 4
43.4
GPT-5
60.7
Full benchmark table
| Benchmark | Claude Opus 4 | GPT-5 |
|---|---|---|
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 72.0 | 88.0 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 35.7 | 65.7 |
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. | 8.6 | 9.9 |
DeepResearch Bench DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses. | 49.0 | 55.1 |
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 61.1 | 97.2 |
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. | 4.5 | 32.4 |
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. | 4.2 | 12.5 |
GeoBench GeoBench · tests geographic knowledge and spatial reasoning across countries, landmarks, coordinates, and geopolitical understanding. | 49.0 | 81.0 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 68.3 | 81.6 |
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 | 6.9 |
HLE HLE (Humanity's Last Exam) · crowdsourced expert-level questions designed to be among the hardest possible challenges for AI systems across all domains. | 6.2 | 21.6 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 85.0 | 98.1 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024–2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 64.4 | 91.4 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 50.6 | 48.0 |
SWE-Bench verified | 70.7 | 73.5 |
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. | 67.6 | 65.0 |
VPCT VPCT (Visual Pattern Completion Test) · tests visual reasoning and pattern recognition by having models complete visual sequences and transformations. | 7.0 | 49.0 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 43.4 | 60.7 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| Claude Opus 4 | $15.00 | $75.00 | 200K tokens (~100 books) | $300.00 |
| GPT-5 | $1.25 | $10.00 | 400K tokens (~200 books) | $34.38 |