Compare · ModelsLive · 2 picked · head to head
Claude Opus 4.7 vs GPT-5.4
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5.4 wins on 14/26 benchmarks
GPT-5.4 wins 14 of 26 shared benchmarks. Leads in agentic · reasoning · knowledge.
Category leads
agentic·GPT-5.4reasoning·GPT-5.4arena·Claude Opus 4.7knowledge·GPT-5.4general·GPT-5.4math·GPT-5.4coding·Claude Opus 4.7
Hype vs Reality
Attention vs performance
Claude Opus 4.7
#124 by perf·#6 by attention
GPT-5.4
#110 by perf·#9 by attention
Best value
GPT-5.4
1.8x better value than Claude Opus 4.7
Claude Opus 4.7
3.4 pts/$
$15.00/M
GPT-5.4
6.0 pts/$
$8.75/M
Vendor risk
Who is behind the model
Anthropic
$965.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
26 benchmarks · 2 models
Claude Opus 4.7GPT-5.4
APEX-Agents
GPT-5.4 leads by +3.2
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Claude Opus 4.7
49.2
GPT-5.4
52.4
ARC-AGI
GPT-5.4 leads by +0.2
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Opus 4.7
93.5
GPT-5.4
93.7
ARC-AGI-2
Claude Opus 4.7 leads by +1.9
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 4.7
75.8
GPT-5.4
74.0
Chatbot Arena Elo · Coding
Claude Opus 4.7 leads by +160.2
Claude Opus 4.7
1555.3
GPT-5.4
1395.0
Chatbot Arena Elo · Overall
Claude Opus 4.7 leads by +29.3
Claude Opus 4.7
1494.1
GPT-5.4
1464.8
Chess Puzzles
GPT-5.4 leads by +7.4
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Claude Opus 4.7
37.9
GPT-5.4
45.3
Dtbench
Claude Opus 4.7 leads by +0.4
Claude Opus 4.7
91.1
GPT-5.4
90.7
Ebr Bench
GPT-5.4 leads by +6.3
Claude Opus 4.7
19.1
GPT-5.4
25.4
FrontierMath-2025-02-28-Private
GPT-5.4 leads by +3.8
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Claude Opus 4.7
43.8
GPT-5.4
47.6
FrontierMath-Tier-4-2025-07-01-Private
GPT-5.4 leads by +4.2
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 4.7
22.9
GPT-5.4
27.1
FrontierMath-Tier-4-v2-Private
GPT-5.4 leads by +17.3
Claude Opus 4.7
31.7
GPT-5.4
49.0
FrontierMath-Tiers-1-3-v2-Private
GPT-5.4 leads by +8.9
Claude Opus 4.7
73.9
GPT-5.4
82.7
Furniture Assembly
GPT-5.4 leads by +6.0
Claude Opus 4.7
4.8
GPT-5.4
10.7
GPQA diamond
GPT-5.4 leads by +4.4
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Opus 4.7
90.5
GPT-5.4
94.9
GSO-Bench
Claude Opus 4.7 leads by +12.7
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 4.7
44.1
GPT-5.4
31.4
HLE
GPT-5.4 leads by +0.1
HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation.
Claude Opus 4.7
33.0
GPT-5.4
33.0
Lmca
Claude Opus 4.7 leads by +0.3
Claude Opus 4.7
61.4
GPT-5.4
61.1
Mirrorcode
Claude Opus 4.7 leads by +7.8
Claude Opus 4.7
31.1
GPT-5.4
23.3
Mystery Game Puzzles
GPT-5.4 leads by +9.9
Claude Opus 4.7
20.7
GPT-5.4
30.6
OTIS Mock AIME 2024-2025
Claude Opus 4.7 leads by +0.0
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Opus 4.7
97.8
GPT-5.4
97.8
PostTrainBench
Claude Opus 4.7 leads by +4.1
Claude Opus 4.7
24.3
GPT-5.4
20.2
Proofbench
GPT-5.4 leads by +2.0
Claude Opus 4.7
54.0
GPT-5.4
56.0
SimpleQA Verified
Claude Opus 4.7 leads by +6.6
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Claude Opus 4.7
51.7
GPT-5.4
45.1
SWE-Bench verified
Claude Opus 4.7 leads by +6.6
SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. The benchmark remains the most-cited evaluation for code-generation capability.
Claude Opus 4.7
83.5
GPT-5.4
76.9
Terminal Bench
GPT-5.4 leads by +1.6
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency.
Claude Opus 4.7
80.2
GPT-5.4
81.8
WeirdML
Claude Opus 4.7 leads by +19.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude Opus 4.7
76.4
GPT-5.4
57.4
Full benchmark table
| Benchmark | Claude Opus 4.7 | GPT-5.4 |
|---|---|---|
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 49.2 | 52.4 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 93.5 | 93.7 |
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. | 75.8 | 74.0 |
Chatbot Arena Elo · Coding | 1555.3 | 1395.0 |
Chatbot Arena Elo · Overall | 1494.1 | 1464.8 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 37.9 | 45.3 |
Dtbench | 91.1 | 90.7 |
Ebr Bench | 19.1 | 25.4 |
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. | 43.8 | 47.6 |
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. | 22.9 | 27.1 |
FrontierMath-Tier-4-v2-Private | 31.7 | 49.0 |
FrontierMath-Tiers-1-3-v2-Private | 73.9 | 82.7 |
Furniture Assembly | 4.8 | 10.7 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 90.5 | 94.9 |
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. | 44.1 | 31.4 |
HLE HLE (Humanity's Last Exam) · a reasoning benchmark designed to be the hardest public evaluation of AI. Questions span mathematics, physics, philosophy, and logic · curated to be at or beyond the frontier of human expert capability. Tested with and without tool augmentation. | 33.0 | 33.0 |
Lmca | 61.4 | 61.1 |
Mirrorcode | 31.1 | 23.3 |
Mystery Game Puzzles | 20.7 | 30.6 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 97.8 | 97.8 |
PostTrainBench | 24.3 | 20.2 |
Proofbench | 54.0 | 56.0 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 51.7 | 45.1 |
SWE-Bench verified SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. The benchmark remains the most-cited evaluation for code-generation capability. | 83.5 | 76.9 |
Terminal Bench Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. | 80.2 | 81.8 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 76.4 | 57.4 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $5.00 | $25.00 | 1.0M tokens (~500 books) | $100.00 | |
| $2.50 | $15.00 | 1.1M tokens (~525 books) | $56.25 |