Claude Sonnet 4.5 vs Claude Sonnet 4.6

Verdict

Claude Sonnet 4.6 leads 14–1 across 15 shared benchmarks.

Claude Sonnet 4.5 1 · Claude Sonnet 4.6 14 · higher isn't always better — check the caveats below.

Per-benchmark head-to-head

Reported scores — protocols may differ. How we compare

Sort

Claude Sonnet 4.5Claude Sonnet 4.6

63.67
BARC-AGI
86.5
13.61
AARC-AGI-2
60.42
7.41
AChess Puzzles
8.46
1.14
ACritPt
3.14
52.6
BDeepResearch Bench
54.87
76.43
BGPQA diamond
83.16
8.57
AMystery Game Puzzles
7.47
62.9
BOSWorld
72.1
77.76
AOTIS Mock AIME 2024-2025
85.78
9.94
APostTrainBench
16.42
19
AProofBench
45
23.6
ASimpleQA Verified
29
71.28
BSWE-Bench verified
75.21
46.5
BTerminal Bench
53.4
47.71
AWeirdML
66.07

Capabilities

Reasoning

Reasoning: Claude Sonnet 4.6 leads 4–0 across 4 shared reasoning benchmarks (largest gap: ARC-AGI, 86.5 vs 63.67 — protocols undisclosed).

4 reproduced4 unverified

Unknown

Unknown: Evenly split 1–1 across 2 shared unknown benchmarks (largest gap: ProofBench, 45 vs 19 — single evaluator).

2 reproduced2 unverified

A higher score does not always mean a better model. Each number is task performance under a disclosed harness. Of the 30 measurements across 15 shared benchmarks, 30 single-evaluator or undisclosed-protocol reports; 40% 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 Sonnet 4.6 leads 14–1 across 15 shared benchmarks. Claude Sonnet 4.5 leads 1 benchmark and Claude Sonnet 4.6 leads 14. Higher isn't always better — see the integrity caveats on each benchmark.

Claude Sonnet 4.5 and Claude Sonnet 4.6 have 15 benchmarks in common in our data — those are the only rows where a direct, apples-to-apples comparison is drawn.

Claude Sonnet 4.6 leads reasoning 4–0.

Follow the record

These models

Same tier