Which is better, Qwen3.5 27B or Qwen3.7 Plus Preview?

Across 52 shared benchmarks tracked on Vector Wire, Qwen3.5 27B leads on 8 and Qwen3.7 Plus Preview leads on 44 — 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 —

Qwen3.5 27B vs Qwen3.7 Plus Preview

AlibabavsAlibaba52 shared benchmarks8–44 head-to-head

Benchmark	Qwen3.5 27B	Qwen3.7 Plus Preview
AA Agentic Index	54.6	20.8
AA Intelligence	34.6	39.4
AA-LCR	72.3	69
AA-Omniscience	-44	2.4
arena_vision	1221	1263
Artificial Analysis Coding Index	34.9	55.9
baby_vision	44.6	70.4
bfcl_v4	68.5	72.9
charxiv_rq	79.5	85.9
Claw Eval (pass@3)	46.2	62.7
critpt	0.9	9.1
DeepPlanning	22.6	62.3
ERQA	60.5	69.8
gdpval	33	22.2
GPQA Diamond	85.8	90.3
HLE	48.5	35.6
hmmt_feb_2026	84.3	92.9
IFBench	76.5	79.1
ifeval	95	94.6
imo_answer_bench	79.9	86
include	81.6	83
LiveCodeBench v6	80.7	89.6
LVBench	73.6	76.2
mathvision	86	90.3
MCP Atlas	68.4	73.2
mcpmark	36.3	58.7
MLVU	85.9	87.4
mmlu_prox	82.2	85.4
mmlu_redux	93.2	94.5
MMLU-Pro	86.1	88.5
mmmlu	85.9	89
MMMU-Pro	75	80.5
nl2repo	27.3	41.1
OmniDocBench 1.5	88.9	91.4
OmniScience Accuracy	20.7	22.5
OmniScience Non-Hallucination	24.9	74.5
OSWorld-Verified	56.2	73.3
QwenClawBench	52.2	61.8
RealWorldQA	83.7	86.9
scicode	39.5	51.3
screenspot_pro_no_tools	70.3	79
SimpleVQA	56	0.8
supergpqa	65.6	71.4
SWE-bench Multilingual	69.3	75.8
SWE-bench Pro	51.2	57.6
SWE-bench Verified	75	77.7
Terminal-Bench 2.0	41.6	70.3
Terminal-Bench Hard	32.6	47
VideoMMMU	82.3	85.4
VITA-Bench	41.9	45.6
τ²-Bench Telecom (AA run)	93.9	93
τ³-Bench	68.4	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.