Claude Opus 4.6 vs GPT-5.3 Codex — Who Actually Wins? Real Benchmark Data (2026)

head to head

Claude Opus 4.6 vs GPT-5.3 Codex

5 shared benchmarks

4

wins

0

ties

1

wins

91.3%

GPQA Diamond

91.5%

53.1%

Humanity's Last Exam

39.9%

72.7%

OSWorld

64.0%

81.42%

SWE-bench Verified

80.0%

81.8%

Terminal-Bench 2.0

77.3%