Benchmark · Knowledge

SWE-Bench Pro (Public)

Updated 2026-01-14
Models tested
2
Top score
45.9
Claude Opus 4.5
Median
43.5
min 41.0
Top-5 spread
σ 2.4
Competitive
SWE-BENCH PRO (PUBLIC) \u00B7 TOP 20255075100#1Claude Opus 4.545.9#2GPT-5.2-Codex41.0benchgecko.ai/benchmark/seal-swe-bench-pro-public

2 models tested · sorted by score

#ModelScore
1Anthropic logoClaude Opus 4.545.9
2OpenAI logoGPT-5.2-Codex41.0

Same category · related evaluations