Which is better, GPT-5.5 or Kimi K2.6?

Across 56 shared benchmarks tracked on Vector Wire, GPT-5.5 leads on 47 and Kimi K2.6 leads on 8 — GPT-5.5 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

GPT-5.5 vs Kimi K2.6

OpenAIvsMoonshot56 shared benchmarks47–8 head-to-head

Benchmark	GPT-5.5	Kimi K2.6
AA Agentic Index	47.4	58.7
AA Intelligence	56.3	45.1
AA-LCR	79	76.7
AA-Omniscience	20.5	6.4
aime_2026	100	96.4
apex_agents	38.5	27.9
apexAgents	37.7	28.5
arena_elo	1482	1461
arena_text_factuality	1482	1455
arena_vision	1283	1264
Artificial Analysis Coding Index	74.9	61.8
browsecomp	84.4	86.3
coding_arena_elo	1457	1509
critpt	27.1	8
Finance Agent v2	51.8	44.9
frontiermath_tier_4	35.4	14.6
gdpval	49.5	41.4
GDPval-AA (Elo)	1769	1482
GDPval-AA v2	1509	1190
GPQA Diamond	93.6	91.1
HiL-Bench	39.7	18.7
HLE	52.2	37.5
HLE (with tools)	52.2	54
hmmt_feb_2026	98.5	94.7
IFBench	75.9	76
itbenchSre	45.8	31.2
livebench	80.7	72.2
livebench_agentic_coding	54	46.9
livebench_coding	82.2	78.6
livebench_data_analysis	81.6	65.1
livebench_instruction_following	70.7	64.4
livebench_language	87.4	75.1
livebench_math	95.9	84.3
livebench_reasoning	89.7	79.4
MCP Atlas	82.8	68.1
MMMU-Pro	83.2	80.1
OmniScience Accuracy	58	32.8
OmniScience Non-Hallucination	12	60.7
OSWorld-Verified	78.7	73.1
Prophet Arena	0.1	0.1
scicode	56.1	53.5
simpleqa_verified	63.1	38.7
SWE-bench Pro	58.6	58.6
SWE-bench Verified	80.6	80.2
TauBench V3 - Banking	39	23.3
Terminal-Bench 2.0	82.7	66.7
Terminal-Bench 2.1	84.3	65.9
Terminal-Bench Hard	60.6	43.9
toolathlon	55.6	50
usamo_2026	98.2	51.2
vectara_answer_rate	100	99.7
vectara_avg_summary_length	129.6	116.7
vectara_factual_consistency	90.7	89.2
vectara_hallucination_rate ↓	9.3	10.8
τ²-Bench Telecom (AA run)	93.9	95.9
τ³-Bench	31.3	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.