Which is better, Gemini 2.5 Pro or o4-mini?

Across 49 shared benchmarks tracked on Vector Wire, Gemini 2.5 Pro leads on 26 and o4-mini leads on 22 — Gemini 2.5 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 2.5 Pro vs o4-mini

GooglevsOpenAI49 shared benchmarks26–22 head-to-head

Benchmark	Gemini 2.5 Pro	o4-mini
AA Agentic Index	7.2	36.1
AA Intelligence	27	26.1
AA-LCR	66	60
AA-Omniscience	-14.3	-35.7
aider_polyglot	82.2	72
AIME 2025	88.3	92.7
AIR-Bench 2024	73.6	78.5
anthropic_red_team	99.5	98.2
arc_agi_1	41	58.7
ARC-AGI-2	4.9	6.1
arena_vision	1246	1201
Artificial Analysis Coding Index	46.7	25.6
bbq	96.4	94
browsecomp	9.9	51.5
critpt	2.6	0.6
Fortress	54.9	21.5
frontiermath_tier_4	4.2	6.3
gdpval	8.5	25.4
GPQA Diamond	84.4	81.4
harmbench	65.4	97
HLE	22.5	16.5
IFBench	49	68.7
imo_2025	31.6	14.3
LiveCodeBench	82.7	84.5
livecodebench_easy	98.8	98.8
livecodebench_hard	59.4	62.9
livecodebench_medium	90.6	92.2
MathVista	83.9	84.3
MMMU	83.9	81.6
MMMU-Pro	74.9	69.2
multichallenge	53.6	44.9
OmniScience Accuracy	39	24.8
OmniScience Non-Hallucination	12.6	19.5
PropensityBench	79	15.8
scicode	43	46.5
SEAL VISTA	50.8	51.8
simple_safety_tests	97	100
simplebench	62.4	38.7
simpleqa_verified	56	23.9
swe_bench_bash	53.6	45
SWE-bench Verified	67.2	68.1
Terminal-Bench Hard	26.5	15.2
usamo_2025	24.4	19.1
vectara_answer_rate	99.1	99.2
vectara_avg_summary_length	106.4	130.9
vectara_factual_consistency	93	81.4
vectara_hallucination_rate ↓	7	18.6
xstest	98.7	97.4
τ²-Bench Telecom (AA run)	54.1	55.6

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.