Claude Opus 4.5 vs Claude Sonnet 4.5

Verdict

Claude Opus 4.5 leads 26–0 across 26 shared benchmarks.

Claude Opus 4.5 26 · Claude Sonnet 4.5 0 · higher isn't always better — check the caveats below.

Per-benchmark head-to-head

Reported scores — protocols may differ. How we compare

Sort

Claude Opus 4.5Claude Sonnet 4.5

80
BARC-AGI
63.67
37.64
AARC-AGI-2
13.61
9.51
AChess Puzzles
5.3
82
BCybench
60
54.8
BDeepResearch Bench
52.6
83.12
ADTBench
72
14.29
AEBR-bench
2.38
4.88
AFrontierMath-Tier-4-v2-Private
2.44
34.39
AFrontierMath-Tiers-1-3-v2-Private
23.86
45.5
AGDPval
42.5
81.4
BGPQA diamond
76.43
26.5
AGSO-Bench
14.71
21.43
AHLE
9.37
52.31
ALMCA
45.6
74.97
BMETR Time Horizons
67.38
14.08
AMystery Game Puzzles
8.57
66.3
BOSWorld
62.9
86.1
BOTIS Mock AIME 2024-2025
77.76
36
BProofBench
19
3.75
BRemote Labor Index
2.08
54.4
ASimpleBench
45.16
45.7
ASimpleQA Verified
30.7
76.65
BSWE-Bench verified
71.28
63.15
BTerminal Bench
46.52
10
AVPCT
9.7
63.74
AWeirdML
47.71

Capabilities

Reasoning

Reasoning: Claude Opus 4.5 leads 6–0 across 6 shared reasoning benchmarks (largest gap: ARC-AGI, 80 vs 63.67 — protocols undisclosed).

2 reproduced10 unverified

Unknown

Unknown: Claude Opus 4.5 leads 6–0 across 6 shared unknown benchmarks (largest gap: ProofBench, 36 vs 19 — single evaluator).

4 reproduced8 unverified

Agentic

Agentic: Claude Opus 4.5 leads 3–0 across 3 shared agentic benchmarks (largest gap: Cybench, 82 vs 60 — single evaluator).

2 self-reported4 unverified

Math

Math: Claude Opus 4.5 leads 2–0 across 2 shared math benchmarks (largest gap: OTIS Mock AIME 2024-2025, 86.1 vs 77.76 — single evaluator).

4 reproduced

A higher score does not always mean a better model. Each number is task performance under a disclosed harness. Of the 52 measurements across 26 shared benchmarks, 52 single-evaluator or undisclosed-protocol reports; 27% are independently reproduced. Claude Sonnet 4.5 pricing is cross-check disputed — weigh cost claims with the same caution as benchmark wins. Check any benchmark's integrity grade before reading a win as decisive.

Frequently asked questions

Claude Opus 4.5 leads 26–0 across 26 shared benchmarks. Claude Opus 4.5 leads 26 benchmarks and Claude Sonnet 4.5 leads 0. Higher isn't always better — see the integrity caveats on each benchmark.

Claude Opus 4.5 and Claude Sonnet 4.5 have 26 benchmarks in common in our data — those are the only rows where a direct, apples-to-apples comparison is drawn.

Claude Opus 4.5 leads reasoning 6–0. Claude Opus 4.5 leads unknown 6–0.

Follow the record

These models

Same tier