Which is better, Gemini 2.5 Pro or GPT-5.1?

Across 46 shared benchmarks tracked on Vector Wire, Gemini 2.5 Pro leads on 10 and GPT-5.1 leads on 36 — GPT-5.1 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 2.5 Pro vs GPT-5.1

GooglevsOpenAI46 shared benchmarks10–36 head-to-head

Benchmark	Gemini 2.5 Pro	GPT-5.1
AA Agentic Index	7.2	32.2
AA Intelligence	27	37.5
AA-LCR	66	76.7
AA-Omniscience	-14.3	5.4
AIME 2025	88.3	94.2
arc_agi_1	41	72.8
ARC-AGI-2	4.9	17.6
arena_elo	1446	1439
arena_text_factuality	1462	1445
arena_vision	1246	1250
Artificial Analysis Coding Index	46.7	49.4
browsecomp	9.9	50.8
coding_arena_elo	1224	1421
critpt	2.6	4.9
Fortress	54.9	25.7
frontiermath_tier_4	4.2	12.5
gdpval	8.5	25
GPQA Diamond	84.4	88.1
HLE	22.5	28.5
HLE (with tools)	28.4	42.7
hmmt_feb_2025	82.5	96.3
hmmt_nov_2025	80	91.7
IFBench	49	72.9
matharena_visual_math_overall	77.2	76.9
MMLU-Pro	86	87
MMMU	83.9	85.4
MMMU-Pro	74.9	75.5
multichallenge	53.6	63.4
OmniScience Accuracy	39	37.7
OmniScience Non-Hallucination	12.6	48.1
scicode	43	43.3
SEAL VISTA	50.8	43.8
simplebench	62.4	53.2
simpleqa_verified	56	48.9
swe_bench_bash	53.6	66
SWE-bench Verified	67.2	76.3
TauBench V3 - Banking	9.7	15.9
Terminal-Bench 2.1	28.5	52.4
Terminal-Bench Hard	26.5	45.5
vectara_answer_rate	99.1	100
vectara_avg_summary_length	106.4	254.4
vectara_factual_consistency	93	89.1
vectara_hallucination_rate ↓	7	12.1
τ²-Bench	54	82.7
τ²-Bench Telecom (AA run)	54.1	81.9
τ³-Bench	9.3	14

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.