Which is better, Gemini 3 Pro or GPT-5.1?

Across 49 shared benchmarks tracked on Vector Wire, Gemini 3 Pro leads on 41 and GPT-5.1 leads on 8 — Gemini 3 Pro leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 3 Pro vs GPT-5.1

GooglevsOpenAI49 shared benchmarks41–8 head-to-head

Benchmark	Gemini 3 Pro	GPT-5.1
AA Agentic Index	52	32.2
AA Intelligence	40.6	37.5
AA-LCR	73	76.7
AA-Omniscience	15.3	5.4
AIME 2025	100	94.2
arc_agi_1	75	72.8
ARC-AGI-2	54	17.6
ARC-AGI-2 (Verified)	31.1	17.6
arena_elo	1485	1439
arena_text_factuality	1481	1445
arena_vision	1290	1250
Artificial Analysis Coding Index	46.5	49.4
browsecomp	59.2	50.8
coding_arena_elo	1438	1421
critpt	9.1	4.9
Fortress	41.7	25.7
frontiermath_tier_4	18.8	12.5
gdpval	34.2	25
GPQA Diamond	91.9	88.1
HLE	45.8	28.5
HLE (with tools)	45.8	42.7
hmmt_feb_2025	97.5	96.3
hmmt_nov_2025	93.3	91.7
IFBench	70.4	72.9
LiveCodeBench v6	90.7	87
matharena_visual_math_overall	84.2	76.9
MCP Atlas	70.3	50.1
MMLU-Pro	90.1	87
mmmlu	91.8	91
MMMU-Pro	81	75.5
multichallenge	65.7	63.4
MultiNRC	59	49
OmniScience Accuracy	55.8	37.7
OmniScience Non-Hallucination	10	48.1
putnam_2025	75.8	48.3
scicode	56.1	43.3
SEAL VISTA	51.5	43.8
simplebench	76.4	53.2
simpleqa_verified	72.9	48.9
swe_bench_bash	74.2	66
SWE-bench Verified	78	76.3
Terminal-Bench 2.0	54.2	47.6
Terminal-Bench Hard	56.9	45.5
vectara_answer_rate	99.4	100
vectara_avg_summary_length	101.9	254.4
vectara_factual_consistency	86.4	89.1
vectara_hallucination_rate ↓	13.6	12.1
τ²-Bench	98	82.7
τ²-Bench Telecom (AA run)	98	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.