Compare · ModelsLive · 2 picked · head to head
Claude Opus 4.8 vs Claude Sonnet 4.5
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Claude Opus 4.8 wins on 20/21 benchmarks
Claude Opus 4.8 wins 20 of 21 shared benchmarks. Leads in reasoning · knowledge · general.
Category leads
reasoning·Claude Opus 4.8knowledge·Claude Opus 4.8general·Claude Opus 4.8math·Claude Opus 4.8coding·Claude Opus 4.8
Hype vs Reality
Attention vs performance
Claude Opus 4.8
#75 by perf·#9 by attention
Claude Sonnet 4.5
#207 by perf·#10 by attention
Best value
Claude Sonnet 4.5
1.1x better value than Claude Opus 4.8
Claude Opus 4.8
3.7 pts/$
$15.00/M
Claude Sonnet 4.5
4.2 pts/$
$9.00/M
Vendor risk
Who is behind the model
Anthropic
$965.0B·Tier 1
Anthropic
$965.0B·Tier 1
Head to head
21 benchmarks · 2 models
Claude Opus 4.8Claude Sonnet 4.5
ARC-AGI
Claude Opus 4.8 leads by +28.8
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Opus 4.8
92.5
Claude Sonnet 4.5
63.7
ARC-AGI-2
Claude Opus 4.8 leads by +58.5
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
72.1
Claude Sonnet 4.5
13.6
Chess Puzzles
Claude Opus 4.8 leads by +23.1
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Claude Opus 4.8
30.6
Claude Sonnet 4.5
7.4
DeepResearch Bench
Claude Sonnet 4.5 leads by +2.4
DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses.
Claude Opus 4.8
50.2
Claude Sonnet 4.5
52.6
Dtbench
Claude Opus 4.8 leads by +19.5
Claude Opus 4.8
91.5
Claude Sonnet 4.5
72.0
Ebr Bench
Claude Opus 4.8 leads by +26.2
Claude Opus 4.8
28.6
Claude Sonnet 4.5
2.4
FrontierMath-2025-02-28-Private
Claude Opus 4.8 leads by +32.0
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Claude Opus 4.8
47.2
Claude Sonnet 4.5
15.2
FrontierMath-Tier-4-2025-07-01-Private
Claude Opus 4.8 leads by +27.1
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 4.8
31.3
Claude Sonnet 4.5
4.2
FrontierMath-Tier-4-v2-Private
Claude Opus 4.8 leads by +53.7
Claude Opus 4.8
56.1
Claude Sonnet 4.5
2.4
FrontierMath-Tiers-1-3-v2-Private
Claude Opus 4.8 leads by +56.1
Claude Opus 4.8
80.0
Claude Sonnet 4.5
23.9
GPQA diamond
Claude Opus 4.8 leads by +11.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Opus 4.8
88.0
Claude Sonnet 4.5
76.4
GSO-Bench
Claude Opus 4.8 leads by +32.4
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 4.8
47.1
Claude Sonnet 4.5
14.7
Lmca
Claude Opus 4.8 leads by +22.1
Claude Opus 4.8
67.7
Claude Sonnet 4.5
45.6
Mystery Game Puzzles
Claude Opus 4.8 leads by +20.9
Claude Opus 4.8
29.5
Claude Sonnet 4.5
8.6
OTIS Mock AIME 2024-2025
Claude Opus 4.8 leads by +20.6
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Opus 4.8
98.3
Claude Sonnet 4.5
77.8
PostTrainBench
Claude Opus 4.8 leads by +23.9
Claude Opus 4.8
33.8
Claude Sonnet 4.5
9.9
Proofbench
Claude Opus 4.8 leads by +50.0
Claude Opus 4.8
69.0
Claude Sonnet 4.5
19.0
Remote Labor Index
Claude Opus 4.8 leads by +6.3
Claude Opus 4.8
8.3
Claude Sonnet 4.5
2.1
SimpleBench
Claude Opus 4.8 leads by +12.6
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Claude Opus 4.8
57.8
Claude Sonnet 4.5
45.2
SimpleQA Verified
Claude Opus 4.8 leads by +22.3
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Claude Opus 4.8
53.0
Claude Sonnet 4.5
30.7
WeirdML
Claude Opus 4.8 leads by +35.2
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude Opus 4.8
82.9
Claude Sonnet 4.5
47.7
Full benchmark table
| Benchmark | Claude Opus 4.8 | Claude Sonnet 4.5 |
|---|---|---|
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 92.5 | 63.7 |
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. | 72.1 | 13.6 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 30.6 | 7.4 |
DeepResearch Bench DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses. | 50.2 | 52.6 |
Dtbench | 91.5 | 72.0 |
Ebr Bench | 28.6 | 2.4 |
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. | 47.2 | 15.2 |
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. | 31.3 | 4.2 |
FrontierMath-Tier-4-v2-Private | 56.1 | 2.4 |
FrontierMath-Tiers-1-3-v2-Private | 80.0 | 23.9 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 88.0 | 76.4 |
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. | 47.1 | 14.7 |
Lmca | 67.7 | 45.6 |
Mystery Game Puzzles | 29.5 | 8.6 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 98.3 | 77.8 |
PostTrainBench | 33.8 | 9.9 |
Proofbench | 69.0 | 19.0 |
Remote Labor Index | 8.3 | 2.1 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 57.8 | 45.2 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 53.0 | 30.7 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 82.9 | 47.7 |
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 | |
| $3.00 | $15.00 | 1.0M tokens (~500 books) | $60.00 |