Which is better, GPT-5.2 or Qwen3.6 Plus?

Across 53 shared benchmarks tracked on Vector Wire, GPT-5.2 leads on 35 and Qwen3.6 Plus leads on 18 — 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 Qwen3.6 Plus

OpenAIvsAlibaba53 shared benchmarks35–18 head-to-head

Benchmark	GPT-5.2	Qwen3.6 Plus
AA Agentic Index	60.2	29
AA Intelligence	43.3	40.5
AA-LCR	79.3	72.3
AA-Omniscience	0.3	2.6
aime_2026	98.3	95.3
arena_elo	1437	1444
arena_text_factuality	1442	1438
Artificial Analysis Coding Index	48.7	54.5
coding_arena_elo	1418	1460
critpt	11.6	2.9
frontiermath_tier_4	18.8	8.3
gdpval	48.3	32
GPQA Diamond	92.4	90.4
HLE	45.5	28.8
HLE (with tools)	45.5	50.6
HMMT 2025	99.4	94.6
hmmt_feb_2026	97	87.8
hmmt_nov_2025	99.2	94.6
IFBench	75.4	75.2
imo_answer_bench	86.3	83.8
livebench	74.8	70.8
livebench_agentic_coding	50.3	41.4
livebench_coding	76.1	78.2
livebench_data_analysis	78.2	69.9
livebench_instruction_following	61.8	58.3
livebench_language	79.8	75
livebench_math	93.2	83.7
livebench_reasoning	83.2	75.8
longbench_v2	54.5	62
mathvision	83	88
MCP Atlas	68	74.1
MMLU-Pro	87	88.5
mmmlu	89.6	89.5
MMMU-Pro	79.5	78.8
OmniDocBench 1.5	85.7	91.2
OmniScience Accuracy	44.3	26.4
OmniScience Non-Hallucination	38.4	68
scicode	52.1	40.7
screenspot_pro_no_tools	86.3	68.2
simpleqa_verified	38.9	49.1
SimpleVQA	55.8	0.7
SWE-bench Multilingual	72	73.8
SWE-bench Pro	55.6	56.6
SWE-bench Verified	80	78.8
TauBench V3 - Banking	11.1	20.8
Terminal-Bench 2.0	54	61.6
Terminal-Bench Hard	62.2	43.9
tool_decathlon	46.3	39.8
toolathlon	46.3	39.8
Video-MME	86	84.2
VideoMMMU	85.9	84
τ²-Bench Telecom (AA run)	98.7	97.7
τ³-Bench	12	70.7

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.