Model	SWE-bench Verified % resolved	GPQA Diamond accuracy	LiveCodeBench Pro Codeforces Elo	Berkeley Function Calling Leaderboard accuracy	LiveBench score	Terminal-Bench 2.1 task success	SWE-bench Pro % resolved	DeepSWE % resolved	Humanity's Last Exam accuracy	MMMU-Pro accuracy	LMArena source-defined rating	ARC-AGI-3 accuracy	ARC-AGI-2 accuracy	FrontierMath accuracy	AIME (OTIS Mock) accuracy	SimpleQA Verified accuracy
GPT-4o GPT	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—
GPT-4.1 GPT	—	—	—	53.96%	—	—	—	—	—	—	—	—	—	—	—	—
o3 o-series	62.3%	—	—	63.05%	—	—	—	—	—	76.4%	—	—	6.53%	18.69%	—	53.0%
o4-mini (high) o-series	—	—	2092	—	—	—	—	—	—	—	—	—	6.11%	—	—	—
o4-mini o-series	—	—	—	53.24%	—	—	—	—	—	—	—	—	—	24.83%	—	—
GPT-OSS-120B GPT OSS	—	—	1299	—	—	—	16.20%	—	—	—	—	—	—	—	88.9%	—
GPT-5 GPT	—	—	—	—	—	—	—	—	—	—	—	—	—	32.41%	—	—
GPT-5 high GPT	73.6%	86.2%	2176	—	—	—	41.78%	—	25.32%	78.4%	—	—	—	—	91.4%	50.6%
GPT-5 mini GPT	64.7%	—	—	55.46%	—	—	—	—	—	—	—	—	—	—	—	—
GPT-5.1 GPT	68.0%	87.6%	2269	—	—	—	—	—	27.2%	79.0%	—	—	—	31.03%	88.6%	48.9%
GPT-5.2 GPT	73.8%	91.4%	2393	55.87%	74.84%	—	29.94%	—	29.9%	80.4%	—	—	—	40.7%	96.1%	—
GPT-5.2-Codex GPT	—	—	—	—	—	—	41.04%	—	—	—	—	—	—	—	—	—
GPT-5.3-Codex GPT	74.8%	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—
GPT-5.4 GPT	76.9%	—	—	—	—	—	—	55.53%	40.28%	82.1%	1472	0.21%	—	47.6%	—	—
GPT-5.4 xHigh GPT	—	93.3%	—	—	80.28%	—	59.10%	—	—	—	—	—	73.95%	—	95.3%	44.8%
GPT-5.4 Pro GPT	—	94.6%	—	—	—	—	—	—	—	—	—	—	83.33%	50.0%	—	47.8%
GPT-5.5 GPT	80.6%	94.0%	—	—	80.71%	—	—	70.05%	43.56%	—	1463	—	85%	51.7%	100.0%	63.1%
GPT-5.5 high GPT	—	—	—	—	—	—	—	—	—	—	1468	0.43%	—	—	—	—
GPT-5.5 Pro GPT	—	93.9%	—	—	—	—	—	—	—	—	—	—	84.58%	52.4%	100.0%	64.5%

Scores are not normalised across benchmarks. Each column uses its own metric. Compare columns independently.