Which is better, Claude 4 Opus or GPT-4o?

Across 55 shared benchmarks tracked on Vector Wire, Claude 4 Opus leads on 47 and GPT-4o leads on 8 — Claude 4 Opus leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude 4 Opus vs GPT-4o

AnthropicvsOpenAI55 shared benchmarks47–8 head-to-head

Benchmark	Claude 4 Opus	GPT-4o
AA Intelligence	31.7	11.2
AA-LCR	40	55
aider_polyglot	70.7	30.7
AIME 2024	76	9.3
AIME 2025	75.5	11.7
AIR-Bench 2024	85.7	62.4
anthropic_red_team	98.9	99.1
arc_agi_1	27	4.5
ARC-AGI-2	8.6	0
arena_vision	1207	1162
Artificial Analysis Coding Index	34	24.2
bbq	99.3	95.1
CNMO 2024	57.6	10.8
Fortress	27.6	47.2
GPQA Diamond	79.6	70.1
harmbench	91.7	82.9
HLE	12.3	5.3
IFBench	53.7	36
ifeval	87.4	84.3
LiveCodeBench	70.4	38.3
LiveCodeBench v6	47.4	30.9
livecodebench_easy	99.1	82.5
livecodebench_hard	33.1	4.5
livecodebench_medium	80.4	32.1
longbench_v2	55.6	48.1
math_500_em	98.2	74.6
metr_hcast	61.7	25
metr_re_bench	20	0
metr_swaa	99.8	99.2
mmlu	92.9	88.1
mmlu_redux	94.2	88
MMLU-Pro	86.6	74.7
mmmlu	88.8	81.4
MMMU	73.7	72.2
multichallenge	58.6	40.3
MultiNRC	33.9	12.4
scicode	40.9	33.4
SEAL VISTA	47	34.9
simple_safety_tests	100	98.5
simplebench	58.8	17.8
simpleqa	22.8	39.4
supergpqa	56.5	42.4
swe_bench_bash	67.6	21.6
SWE-bench Verified	79.4	38.8
TAU-bench (airline)	59.6	42.8
TAU-bench (retail)	81.4	60.3
Tau2 airline	60	45.5
Tau2 retail	81.8	63.4
Terminal-Bench Hard	31.1	8.3
vectara_answer_rate	91	93.8
vectara_avg_summary_length	123.2	86.6
vectara_factual_consistency	88	90.4
vectara_hallucination_rate ↓	12	9.6
xstest	97	97.3
τ²-Bench Telecom (AA run)	73.4	28.9

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.