GPT-5.3 Codex vs GPT-5.4 Mini — Who Actually Wins? Real Benchmark Data (2026)

head to head

GPT-5.3 Codex vs GPT-5.4 Mini

4 shared benchmarks

3

wins

0

ties

1

wins

91.5%

GPQA Diamond

88.0%

64.0%

OSWorld

72.1%

56.8%

SWE-bench Pro

54.4%

77.3%

Terminal-Bench 2.0

60.0%