Which is better, Gemini 2.5 Pro or Kimi K2.5?

Across 53 shared benchmarks tracked on Vector Wire, Gemini 2.5 Pro leads on 10 and Kimi K2.5 leads on 43 — Kimi K2.5 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 Kimi K2.5

GooglevsMoonshot53 shared benchmarks10–43 head-to-head

Benchmark	Gemini 2.5 Pro	Kimi K2.5
AA Agentic Index	7.2	52.8
AA Intelligence	27	36
AA-LCR	66	73
AA-Omniscience	-14.3	-7.3
AIME 2025	88.3	96.1
arc_agi_1	41	65.3
ARC-AGI-2	4.9	11.8
arena_elo	1446	1450
arena_text_factuality	1462	1445
arena_vision	1246	1247
Artificial Analysis Coding Index	46.7	46.8
browsecomp	9.9	74.9
browsecomp_zh	32.2	62.3
coding_arena_elo	1224	1436
critpt	2.6	3.1
Fortress	54.9	41.1
frontiermath_tier_4	4.2	4.2
gdpval	8.5	38.3
Global-MMLU-Lite	88.6	84
GPQA Diamond	84.4	87.9
HLE	22.5	50.2
HLE (with tools)	28.4	51.8
HMMT 2025	65.7	95.4
hmmt_feb_2025	82.5	95.4
hmmt_nov_2025	80	91.1
IFBench	49	70.2
matharena_visual_math_overall	77.2	80.6
mathvision	73.3	84.2
MathVista	83.9	90.1
MMLU-Pro	86	87.1
MMMU-Pro	74.9	78.5
multichallenge	53.6	61.4
OCRBench	85.9	92.3
OmniScience Accuracy	39	35.2
OmniScience Non-Hallucination	12.6	50
scicode	43	49
SEAL VISTA	50.8	41.9
simplebench	62.4	46.8
simpleqa_verified	56	36.9
swe_bench_bash	53.6	70.8
SWE-bench Verified	67.2	76.8
TauBench V3 - Banking	9.7	14.2
Terminal-Bench 2.1	28.5	45.7
Terminal-Bench Hard	26.5	34.8
vectara_answer_rate	99.1	92.2
vectara_avg_summary_length	106.4	112
vectara_factual_consistency	93	85.8
vectara_hallucination_rate ↓	7	14.2
Video-MME	84.8	87.4
xbench_deepsearch	56	76.7
τ²-Bench	54	85.4
τ²-Bench Telecom (AA run)	54.1	95.9
τ³-Bench	9.3	66

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.