Which is better, GPT-5.2 or Qwen3.5 27B?

Across 51 shared benchmarks tracked on Vector Wire, GPT-5.2 leads on 40 and Qwen3.5 27B leads on 11 — GPT-5.2 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

GPT-5.2 vs Qwen3.5 27B

OpenAIvsAlibaba51 shared benchmarks40–11 head-to-head

Benchmark	GPT-5.2	Qwen3.5 27B
AA Agentic Index	60.2	54.6
AA Intelligence	43.3	34.6
AA-LCR	79.3	72.3
AA-Omniscience	0.3	-44
aime_2026	98.3	92.6
arena_vision	1245	1221
Artificial Analysis Coding Index	48.7	34.9
browsecomp	65.8	61
browsecomp_zh	76.1	62.1
charxiv_rq	82.1	79.5
coding_arena_elo	1418	1357
critpt	11.6	0.9
gdpval	48.3	33
GPQA Diamond	92.4	85.8
HLE	45.5	48.5
HMMT 2025	99.4	89.8
hmmt_feb_2025	100	92
hmmt_feb_2026	97	84.3
hmmt_nov_2025	99.2	89.8
IFBench	75.4	76.5
imo_answer_bench	86.3	79.9
longbench_v2	54.5	60.6
mathvision	83	86
MathVista	82.8	87.8
MCP Atlas	68	68.4
MMLU-Pro	87	86.1
mmmlu	89.6	85.9
MMMU-Pro	79.5	75
MMVU	80.8	73.3
OCRBench	80.7	89.4
OmniDocBench 1.5	85.7	88.9
OmniScience Accuracy	44.3	20.7
OmniScience Non-Hallucination	38.4	24.9
scicode	52.1	39.5
screenspot_pro_no_tools	86.3	70.3
Seal-0	45	47.2
SimpleVQA	55.8	56
SWE-bench Multilingual	72	69.3
SWE-bench Pro	55.6	51.2
SWE-bench Verified	80	75
Terminal-Bench 2.0	54	41.6
Terminal-Bench Hard	62.2	32.6
tool_decathlon	46.3	31.5
vectara_answer_rate	100	99.8
vectara_avg_summary_length	186.3	94.4
vectara_factual_consistency	89.2	87.9
vectara_hallucination_rate ↓	10.8	12.1
VideoMMMU	85.9	82.3
ZeroBench	9	0.1
τ²-Bench Telecom (AA run)	98.7	93.9
τ³-Bench	12	68.4

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.