Which is better, Qwen3.6 27B or Qwen3.7 Plus Preview?

Across 52 shared benchmarks tracked on Vector Wire, Qwen3.6 27B leads on 6 and Qwen3.7 Plus Preview leads on 46 — Qwen3.7 Plus Preview leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Qwen3.6 27B vs Qwen3.7 Plus Preview

AlibabavsAlibaba52 shared benchmarks6–46 head-to-head

Benchmark	Qwen3.6 27B	Qwen3.7 Plus Preview
AA Agentic Index	27.5	20.8
AA Intelligence	37.7	39.4
AA-LCR	73.3	69
AA-Omniscience	-20	2.4
Agents' Last Exam	27.3	33.6
AndroidWorld	70.3	81
Artificial Analysis Coding Index	53.7	55.9
baby_vision	28.9	70.4
charxiv_rq	78.4	85.9
Claw Eval (pass@3)	60.6	62.7
ClawEval-MM	50.4	60.1
critpt	1.1	9.1
DeepSWE 1.1	13.3	14.2
ERQA	62.5	69.8
gdpval	32	22.2
GPQA Diamond	87.8	90.3
HLE	24	35.6
hmmt_feb_2026	84.3	92.9
IFBench	69.1	79.1
imo_answer_bench	80.8	86
JobBench	21.8	27.6
LiveCodeBench v6	83.9	89.6
mathvision	85.1	90.3
MLVU	86.6	87.4
mmlu_redux	93.5	94.5
MMLU-Pro	86.2	88.5
MMMU-Pro	75.8	80.5
nl2repo	36.2	41.1
OmniDocBench 1.5	89.4	91.4
OmniScience Accuracy	19.6	22.5
OmniScience Non-Hallucination	50.7	74.5
OSWorld-Verified	63.9	73.3
QwenClawBench	53.4	61.8
RealWorldQA	84.1	86.9
RecreationBench	29.8	30.2
scicode	39.8	51.3
SimpleVQA	56.1	0.8
supergpqa	66	71.4
SWE-bench Multilingual	71.3	75.8
SWE-bench Pro	53.5	57.6
SWE-bench Verified	77.2	77.7
SWE-MM	25.7	30
TauBench V3 - Banking	16.7	17.9
Terminal Bench 2.1 (Terminus)	63.4	64
Terminal-Bench 2.0	59.3	70.3
Terminal-Bench 2.1	60.7	61
Terminal-Bench Hard	34.8	47
VideoMMMU	84.4	85.4
Vision2Web	45	42.1
WebArena-Verified	48.8	55.3
τ²-Bench Telecom (AA run)	94.2	93
τ³-Bench	15.3	17.9

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.