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

Across 43 shared benchmarks tracked on Vector Wire, Gemini 3 Flash Preview leads on 11 and Kimi K2.6 leads on 32 — 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 Flash Preview vs Kimi K2.6

GooglevsMoonshot43 shared benchmarks11–32 head-to-head

Benchmark	Gemini 3 Flash Preview	Kimi K2.6
AA Agentic Index	49.7	58.7
AA Intelligence	38.7	45.1
AA-LCR	73	76.7
AA-Omniscience	10.1	6.4
aime_2026	96.7	96.4
apexAgents	27.7	28.5
arena_elo	1473	1461
arena_text_factuality	1452	1455
arena_vision	1260	1264
Artificial Analysis Coding Index	42.6	61.8
charxiv_rq	80.3	86.7
coding_arena_elo	1387	1509
critpt	8.6	8
Finance Agent v2	42.6	44.9
frontiermath_tier_4	4.2	14.6
gdpval	35.2	41.4
GDPval-AA (Elo)	1204	1482
GPQA Diamond	90.4	91.1
HLE	43.5	37.5
hmmt_feb_2026	89.4	94.7
IFBench	78	76
livebench	72.4	72.2
MCP Atlas	62	68.1
MMMU-Pro	81.2	80.1
OmniScience Accuracy	53.4	32.8
OmniScience Non-Hallucination	7.6	60.7
OSWorld-Verified	65.1	73.1
scicode	50.6	53.5
simpleqa_verified	67.4	38.7
SWE-bench Multilingual	72.7	76.7
SWE-bench Pro	49.6	58.6
SWE-bench Verified	78	80.2
TauBench V3 - Banking	20.8	23.3
Terminal-Bench 2.0	47.6	66.7
Terminal-Bench 2.1	58	65.9
Terminal-Bench Hard	38.6	43.9
toolathlon	49.4	50
vectara_answer_rate	99.8	99.7
vectara_avg_summary_length	90.2	116.7
vectara_factual_consistency	86.5	89.2
vectara_hallucination_rate ↓	13.5	10.8
τ²-Bench Telecom (AA run)	80.4	95.9
τ³-Bench	17.5	20.6

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.