Which is better, Kimi K2.5 or Qwen3.5 122B A10B?

Refresh Models —Deals —Regulatory —Sources —

Questions this page answers

Which is better, Kimi K2.5 or Qwen3.5 122B A10B?: Across 47 shared benchmarks, Kimi K2.5 scores higher on 36 and Qwen3.5 122B A10B on 11. The widest gap is τ³-Bench, where Kimi K2.5 scores 66 against 13.6. Tracked API pricing per million tokens: Kimi K2.5 $0.56 in / $2.94 out, Qwen3.5 122B A10B $0.40 in / $3.20 out.

Kimi K2.5 vs Qwen3.5 122B A10B

Across 47 shared benchmarks, Kimi K2.5 scores higher on 36 and Qwen3.5 122B A10B on 11. The widest gap is τ³-Bench, where Kimi K2.5 scores 66 against 13.6. Tracked API pricing per million tokens: Kimi K2.5 $0.56 in / $2.94 out, Qwen3.5 122B A10B $0.40 in / $3.20 out.

MoonshotvsAlibaba47 shared benchmarks36–11 head-to-head

Benchmark	Kimi K2.5	Qwen3.5 122B A10B
AA Agentic Index	52.8	21.3
AA Intelligence	36	32.8
AA-LCR	73	70.3
AA-Omniscience	-7.3	-41.5
arena_vision	1267	1246
Artificial Analysis Coding Index	46.8	45.7
baby_vision	36.5	40.2
browsecomp	74.9	63.8
browsecomp_zh	62.3	69.9
coding_arena_elo	1436	1358
critpt	3.1	0.9
gdpval	38.3	24.3
GPQA Diamond	87.9	86.6
HLE	50.2	47.5
HMMT 2025	95.4	90.3
IFBench	70.2	76.1
LiveCodeBench v6	85	78.9
longbench_v2	61	60.2
LVBench	75.9	74.4
mathvision	84.2	86.2
MathVista	90.1	87.4
MMLU-Pro	87.1	86.7
MMMU-Pro	78.5	76.9
MMVU	80.4	74.7
multichallenge	61.4	61.5
OCRBench	92.3	92.1
OmniDocBench 1.5	88.8	89.8
OmniScience Accuracy	35.2	24.4
OmniScience Non-Hallucination	50	12.9
OSWorld-Verified	63.3	58
scicode	49	42
Seal-0	57.4	44.1
SimpleVQA	71.2	0.6
SWE-bench Verified	76.8	72
TauBench V3 - Banking	14.2	15.3
Terminal-Bench 2.0	50.8	49.4
Terminal-Bench 2.1	45.7	47.6
Terminal-Bench Hard	34.8	31.1
vectara_answer_rate	92.2	99.8
vectara_avg_summary_length	112	86.4
vectara_factual_consistency	85.8	88.8
vectara_hallucination_rate ↓	14.2	11.2
VideoMMMU	86.6	82
WideSearch	79	60.5
ZeroBench	9	0.1
τ²-Bench Telecom (AA run)	95.9	93.6
τ³-Bench	66	13.6

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. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (moonshot-official, alibaba-official), otherwise the lowest tracked offer.