Which is better, Gemini 2.5 Flash or Gemini 2.5 Pro?

Across 48 shared benchmarks tracked on Vector Wire, Gemini 2.5 Flash leads on 8 and Gemini 2.5 Pro leads on 39 — 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 Flash vs Gemini 2.5 Pro

GooglevsGoogle48 shared benchmarks8–39 head-to-head

Benchmark	Gemini 2.5 Flash	Gemini 2.5 Pro
AA Agentic Index	18.8	7.2
AA Intelligence	17.7	27
AA-LCR	65.7	66
AA-Omniscience	-29.8	-14.3
aider_polyglot	61.9	82.2
AIME 2025	72	88.3
AIR-Bench 2024	68.7	73.6
anthropic_red_team	98.8	99.5
arc_agi_1	33.3	41
ARC-AGI-2	2.5	4.9
arena_text_factuality	1437	1462
arena_vision	1214	1246
Artificial Analysis Coding Index	22.2	46.7
bbq	97.7	96.4
critpt	1.4	2.6
frontiermath_tier_4	4.2	4.2
gdpval	11.9	8.5
Global-MMLU-Lite	88.4	88.6
GPQA Diamond	82.8	84.4
harmbench	62.6	65.4
HLE	12.1	22.5
hmmt_feb_2025	64.2	82.5
IFBench	50.3	49
LiveCodeBench	76.2	82.7
LiveCodeBench (v5)	63.9	75.6
livecodebench_easy	99.1	98.8
livecodebench_hard	48.3	59.4
livecodebench_medium	82.5	90.6
MMMU	79.7	83.9
MMMU-Pro	69.1	74.9
mrcr	32	93
OmniScience Accuracy	26.1	39
OmniScience Non-Hallucination	24.7	12.6
PropensityBench	68	79
scicode	35.9	43
SEAL VISTA	49.1	50.8
simple_safety_tests	98	97
simplebench	41.2	62.4
simpleqa	26.9	50.8
swe_bench_bash	28.7	53.6
SWE-bench Verified	60.4	67.2
Terminal-Bench Hard	13.6	26.5
vectara_answer_rate	99	99.1
vectara_avg_summary_length	101.5	106.4
vectara_factual_consistency	92.2	93
vectara_hallucination_rate ↓	7.8	7
xstest	98.8	98.7
τ²-Bench Telecom (AA run)	31.6	54.1

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.