Which is better, Claude Opus 4.5 or GPT-5.1?

Across 49 shared benchmarks tracked on Vector Wire, Claude Opus 4.5 leads on 27 and GPT-5.1 leads on 21 — Claude Opus 4.5 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 4.5 vs GPT-5.1

AnthropicvsOpenAI49 shared benchmarks27–21 head-to-head

Benchmark	Claude Opus 4.5	GPT-5.1
AA Agentic Index	59.6	32.2
AA Intelligence	41.9	37.5
AA-LCR	76	76.7
AA-Omniscience	14	5.4
AIME 2025	92.8	94.2
arc_agi_1	82	72.8
ARC-AGI-2	37.6	17.6
ARC-AGI-2 (Verified)	37.6	17.6
arena_elo	1473	1439
arena_text_factuality	1458	1445
Artificial Analysis Coding Index	47.8	49.4
browsecomp	37	50.8
coding_arena_elo	1495	1421
critpt	4.6	4.9
Fortress	13.6	25.7
frontiermath_tier_4	4.2	12.5
gdpval	47.3	25
GPQA Diamond	87	88.1
HLE	30.8	28.5
HLE (with tools)	43.4	42.7
hmmt_feb_2025	92.9	96.3
hmmt_nov_2025	93.3	91.7
IFBench	58	72.9
LiveCodeBench v6	84.8	87
MCP Atlas	69.8	50.1
MMLU-Pro	90	87
mmmlu	90.8	91
MMMU	80.7	85.4
MMMU (validation)	80.7	85.4
MMMU-Pro	74	75.5
multichallenge	59	63.4
MultiNRC	48.6	49
OmniScience Accuracy	46.6	37.7
OmniScience Non-Hallucination	39	48.1
scicode	50	43.3
SEAL VISTA	46.4	43.8
simplebench	62	53.2
simpleqa_verified	41.8	48.9
swe_bench_bash	76.8	66
SWE-bench Verified	81.5	76.3
Tau2 retail	88.9	77.9
Terminal-Bench 2.0	59.3	47.6
Terminal-Bench Hard	47	45.5
vectara_answer_rate	98.7	100
vectara_avg_summary_length	114.5	254.4
vectara_factual_consistency	89.1	89.1
vectara_hallucination_rate ↓	10.9	12.1
τ²-Bench	98.2	82.7
τ²-Bench Telecom (AA run)	89.5	81.9

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.