Which is better, Claude Opus 4.8 or Kimi K2.5?

Across 59 shared benchmarks tracked on Vector Wire, Claude Opus 4.8 leads on 50 and Kimi K2.5 leads on 8 — Claude Opus 4.8 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 4.8 vs Kimi K2.5

AnthropicvsMoonshot59 shared benchmarks50–8 head-to-head

Benchmark	Claude Opus 4.8	Kimi K2.5
AA Agentic Index	49.4	52.8
AA Intelligence	57.3	36
AA-LCR	73	73
AA-Omniscience	28.8	-7.3
aime_2026	100	95.8
arc_agi_1	92.5	65.3
ARC-AGI-2	72.1	11.8
arena_elo	1482	1450
arena_text_factuality	1464	1445
arena_vision	1280	1247
Artificial Analysis Coding Index	74.3	46.8
baby_vision_with_python	81.2	40.5
browsecomp	88.5	74.9
charxiv_rq	80.5	78.7
coding_arena_elo	1539	1436
critpt	20.9	3.1
cybergym	83.1	41.3
deepsearchqa_f1	93.1	89
Fortress	18.2	41.1
frontiermath_tier_4	31.3	4.2
gdpval	54.2	38.3
GDPval-AA v2	1593	1009
GPQA Diamond	93.6	87.9
HLE	57.9	50.2
HLE (with tools)	57.9	51.8
hmmt_feb_2026	96.7	87.1
hmmt_nov_2025	96.5	91.1
IFBench	62.2	70.2
imo_answer_bench	83.5	81.8
livebench	77.2	69.1
longbench_v2	69.1	61
matharena_visual_math_overall	81.6	80.6
mathvision	86.7	84.2
MCP Atlas	83.6	64
MMMU-Pro	78.9	78.5
MMVU	79.2	80.4
nl2repo	69.7	32
OmniScience Accuracy	48.8	35.2
OmniScience Non-Hallucination	60.7	50
OSWorld-Verified	83.4	63.3
PaperBench	80.3	63.5
ResearchRubrics	73.5	59.5
scicode	53.5	49
simplebench	64.8	46.8
simpleqa_verified	39.5	36.9
SWE-bench Multilingual	84.4	73
SWE-bench Pro	69.2	53.8
SWE-bench Verified	88.6	76.8
Tau 3 Banking	27.6	14.2
TauBench V3 - Banking	34.2	14.2
Terminal-Bench 2.0	74.6	50.8
Terminal-Bench 2.1	85	45.7
Terminal-Bench Hard	58.3	34.8
tool_decathlon	59.9	27.8
toolathlon	59.9	27.8
Video-MME	86	87.4
WideSearch	72.9	79
τ²-Bench Telecom (AA run)	94.4	95.9
τ³-Bench	27.6	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.