Which is better, Gemini 3 Pro or Kimi K2.6?

Across 47 shared benchmarks tracked on Vector Wire, Gemini 3 Pro leads on 19 and Kimi K2.6 leads on 28 — Kimi K2.6 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 Kimi K2.6

GooglevsMoonshot47 shared benchmarks19–28 head-to-head

Benchmark	Gemini 3 Pro	Kimi K2.6
AA Agentic Index	52	58.7
AA Intelligence	40.6	45.1
AA-LCR	73	76.7
AA-Omniscience	15.3	6.4
aime_2026	91.7	96.4
apexAgents	18.4	28.5
arena_elo	1485	1461
arena_text_factuality	1481	1455
arena_vision	1290	1264
Artificial Analysis Coding Index	46.5	61.8
browsecomp	59.2	86.3
charxiv_rq	81.4	86.7
coding_arena_elo	1438	1509
critpt	9.1	8
deepsearchqa_f1	63.2	92.5
frontiermath_tier_4	18.8	14.6
gdpval	34.2	41.4
GDPval-AA (Elo)	1195	1482
GPQA Diamond	91.9	91.1
HLE	45.8	37.5
HLE (with tools)	45.8	54
hmmt_feb_2026	86.4	94.7
IFBench	70.4	76
imo_answer_bench	83.3	86
livebench	73.4	72.2
LiveCodeBench	92	89.6
LiveCodeBench v6	90.7	89.6
mathvision	86.1	93.2
MCP Atlas	70.3	68.1
MMLU-Pro	90.1	87.1
MMMU-Pro	81	80.1
OmniScience Accuracy	55.8	32.8
OmniScience Non-Hallucination	10	60.7
scicode	56.1	53.5
simpleqa_verified	72.9	38.7
SWE-bench Multilingual	68.7	76.7
SWE-bench Pro	43.3	58.6
SWE-bench Verified	78	80.2
Terminal-Bench 2.0	54.2	66.7
Terminal-Bench Hard	56.9	43.9
toolathlon	36.4	50
vectara_answer_rate	99.4	99.7
vectara_avg_summary_length	101.9	116.7
vectara_factual_consistency	86.4	89.2
vectara_hallucination_rate ↓	13.6	10.8
WideSearch (item-f1)	57	80.8
τ²-Bench Telecom (AA run)	98	95.9

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.