Which is better, Claude Opus 4.6 or GPT-5?

Across 55 shared benchmarks tracked on Vector Wire, Claude Opus 4.6 leads on 42 and GPT-5 leads on 13 — Claude Opus 4.6 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Opus 4.6 vs GPT-5

AnthropicvsOpenAI55 shared benchmarks42–13 head-to-head

Benchmark	Claude Opus 4.6	GPT-5
AA Agentic Index	67.6	49.7
AA Intelligence	44.9	35.3
AA-LCR	74.3	76.3
AA-Omniscience	13.7	-8.7
AIME 2025	99.8	95
AIME25 no tools	95.6	94.6
arc_agi_1	93	56.2
ARC-AGI-2	68.8	7.5
arena_vision	1300	1212
Artificial Analysis Coding Index	48.1	38.9
browsecomp	86.8	54.9
coding_arena_elo	1545	1419
critpt	12.6	5.7
ERQA	40.8	65.7
Fortress	20.5	17
frontiermath_tier_4	22.9	12.5
gdpval	55.9	32.5
GPQA Diamond	91.3	87.3
GSM8K	0.6	-1.8
HLE	53.1	28.5
HLE (with tools)	62.7	35.2
hmmt_nov_2025	96.3	89.2
IFBench	62.5	73.1
imo_answer_bench	75.3	76
LiveCodeBench	88.8	85
LiveCodeBench v6	88.8	87
matharena_visual_math_overall	72.3	78.8
metr_hcast	77	68.4
metr_re_bench	97.4	40
metr_swaa	98.5	100
MMLU-Pro	89.1	87.5
MMMU-Pro	77.3	78.4
multichallenge	56	71.1
MultiNRC	57.1	52.1
ocrbench_v2	48.4	55.5
OmniScience Accuracy	47	40.3
OmniScience Non-Hallucination	37.2	21
scicode	52	43
SEAL VISTA	46.1	49.7
simplebench	67.6	56.7
simpleqa_verified	46.5	50.6
swe_bench_bash	75.6	65
SWE-bench Multilingual	77.8	55.3
SWE-bench Pro	57.3	41.8
SWE-bench Verified	80.8	74.9
Tau2 retail	91.9	81.1
Terminal-Bench 2.0	65.4	35.2
Terminal-Bench Hard	65.4	37.9
vectara_answer_rate	99.8	99.9
vectara_avg_summary_length	137.6	162.7
vectara_factual_consistency	87.8	85.3
vectara_hallucination_rate ↓	12.2	14.7
VisualToolBench	27.5	18.7
τ²-Bench Telecom (AA run)	99.3	86.5
τ³-Bench	72.4	19.6

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.