Which is better, GPT-5.2 or Qwen3.5 35B A3B?

Across 53 shared benchmarks tracked on Vector Wire, GPT-5.2 leads on 42 and Qwen3.5 35B A3B 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 35B A3B

OpenAIvsAlibaba53 shared benchmarks42–11 head-to-head

Benchmark	GPT-5.2	Qwen3.5 35B A3B
AA Agentic Index	60.2	44.1
AA Intelligence	43.3	29.9
AA-LCR	79.3	68.3
AA-Omniscience	0.3	-48.1
aime_2026	98.3	93.3
Artificial Analysis Coding Index	48.7	37
browsecomp	65.8	61
browsecomp_zh	76.1	69.5
charxiv_rq	82.1	77.5
coding_arena_elo	1418	1250
critpt	11.6	0.9
gdpval	48.3	20.3
GPQA (unspecified)	92.4	84.2
GPQA Diamond	92.4	84.5
HLE	45.5	47.4
HMMT 2025	99.4	89.2
hmmt_feb_2025	100	89
hmmt_feb_2026	97	81.8
hmmt_nov_2025	99.2	89.2
IFBench	75.4	72.5
imo_answer_bench	86.3	76.8
longbench_v2	54.5	59
mathvision	83	83.9
MathVista	82.8	86.2
MCP Atlas	68	62.4
MMLU-Pro	87	85.3
mmmlu	89.6	85.2
MMMU-Pro	79.5	75.1
MMVU	80.8	72.3
OCRBench	80.7	91
ocrbench_v2	50.5	65.3
OmniDocBench 1.5	85.7	89.3
OmniScience Accuracy	44.3	20.1
OmniScience Non-Hallucination	38.4	14.6
scicode	52.1	37.7
screenspot_pro_no_tools	86.3	68.6
Seal-0	45	41.4
SimpleVQA	55.8	58.3
SWE-bench Multilingual	72	60.3
SWE-bench Pro	55.6	44.6
SWE-bench Verified	80	70
TauBench V3 - Banking	11.1	4.9
Terminal-Bench 2.0	54	40.5
Terminal-Bench Hard	62.2	26.5
tool_decathlon	46.3	28.7
vectara_answer_rate	100	99.8
vectara_avg_summary_length	186.3	94.9
vectara_factual_consistency	89.2	89.5
vectara_hallucination_rate ↓	10.8	10.5
VideoMMMU	85.9	80.4
ZeroBench	9	0.1
τ²-Bench Telecom (AA run)	98.7	89.2
τ³-Bench	12	68.9

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.