Which is better, Claude Opus 4.6 or Kimi K2.6?

Across 79 shared benchmarks tracked on Vector Wire, Claude Opus 4.6 leads on 50 and Kimi K2.6 leads on 29 — Claude Opus 4.6 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Opus 4.6 vs Kimi K2.6

AnthropicvsMoonshot79 shared benchmarks50–29 head-to-head

Benchmark	Claude Opus 4.6	Kimi K2.6
AA Agentic Index	67.6	58.7
AA Intelligence	44.9	45.1
AA-LCR	74.3	76.7
AA-Omniscience	13.7	6.4
AgentWorldBench - Android	61.7	58.9
AgentWorldBench - MCP	69.9	65.2
AgentWorldBench - OS	70.2	60.8
AgentWorldBench - Overall	57.8	53.4
AgentWorldBench - Search	29.3	27.5
AgentWorldBench - SWE	64.5	58.8
AgentWorldBench - Terminal	57.5	52.5
AgentWorldBench - Web	51.4	50.2
aime_2026	96.7	96.4
Apex (Pass@1)	34.5	24
Apex Shortlist (Pass@1)	85.9	75.5
apexAgents	33	28.5
arena_elo	1497	1461
arena_text_factuality	1487	1455
arena_vision	1300	1264
Artificial Analysis Coding Index	48.1	61.8
baby_vision	14.8	68.5
baby_vision_with_python	38.4	68.5
browsecomp	86.8	86.3
BrowseComp (Agent Swarm)	83.7	86.3
charxiv_rq	69.1	86.7
charxiv_rq_with_python	84.7	86.7
Chinese SimpleQA (C-SimpleQA)	76.4	75.9
coding_arena_elo	1545	1509
critpt	12.6	8
deepsearchqa_accuracy	80.6	83
deepsearchqa_f1	91.3	92.5
frontiermath_tier_4	22.9	14.6
gdpval	55.9	41.4
GDPval-AA (Elo)	1619	1482
GPQA Diamond	91.3	91.1
HiL-Bench	38.3	18.7
HLE	53.1	37.5
HLE (with tools)	62.7	54
hmmt_feb_2026	96.2	94.7
IFBench	62.5	76
imo_answer_bench	75.3	86
livebench	76.3	72.2
livebench_agentic_coding	49	46.9
livebench_coding	78.2	78.6
livebench_data_analysis	69.9	65.1
livebench_instruction_following	63.3	64.4
livebench_language	83.3	75.1
livebench_math	89.3	84.3
livebench_reasoning	88.7	79.4
LiveCodeBench	88.8	89.6
LiveCodeBench v6	88.8	89.6
mathvision	71.2	93.2
mathvision_with_python	84.6	93.2
MCP Atlas	76.8	68.1
MCPAtlas Public (Pass@1)	73.8	66.6
mcpmark	56.7	55.9
MMLU-Pro	89.1	87.1
mmmu_pro_with_python	77.3	80.1
MMMU-Pro	77.3	80.1
OJBench (python)	60.3	60.6
OmniScience Accuracy	47	32.8
OmniScience Non-Hallucination	37.2	60.7
OSWorld-Verified	72.7	73.1
scicode	52	53.5
simpleqa_verified	46.5	38.7
SWE-bench Multilingual	77.8	76.7
SWE-bench Pro	57.3	58.6
SWE-bench Verified	80.8	80.2
Terminal-Bench 2.0	65.4	66.7
Terminal-Bench Hard	65.4	43.9
toolathlon	56.8	50
usamo_2026	66.2	51.2
V* (w/ python)	86.4	96.9
vectara_answer_rate	99.8	99.7
vectara_avg_summary_length	137.6	116.7
vectara_factual_consistency	87.8	89.2
vectara_hallucination_rate ↓	12.2	10.8
τ²-Bench Telecom (AA run)	99.3	95.9
τ³-Bench	72.4	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.