Which is better, GPT-5.2 or Qwen3.7 Max?

Across 44 shared benchmarks tracked on Vector Wire, GPT-5.2 leads on 16 and Qwen3.7 Max leads on 27 — Qwen3.7 Max 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 Qwen3.7 Max

OpenAIvsAlibaba44 shared benchmarks16–27 head-to-head

Benchmark	GPT-5.2	Qwen3.7 Max
AA Agentic Index	60.2	30.9
AA Intelligence	43.3	46.7
AA-LCR	79.3	74.7
AA-Omniscience	0.3	13.5
aime_2026	98.3	97
arena_elo	1437	1474
arena_text_factuality	1442	1479
Artificial Analysis Coding Index	48.7	66
coding_arena_elo	1418	1517
critpt	11.6	13.4
gdpval	48.3	38.5
GPQA Diamond	92.4	92.4
HLE	45.5	41.4
HLE (with tools)	45.5	53.5
hmmt_feb_2026	97	97.1
hmmt_nov_2025	99.2	95
IFBench	75.4	80.5
imo_answer_bench	86.3	90
livebench	74.8	74.3
livebench_agentic_coding	50.3	43.6
livebench_coding	76.1	74.2
livebench_data_analysis	78.2	71.8
livebench_instruction_following	61.8	74
livebench_language	79.8	79.7
livebench_math	93.2	85.3
livebench_reasoning	83.2	83.3
longbench_v2	54.5	65.3
MCP Atlas	68	76.4
MMLU-Pro	87	89.6
mmmlu	89.6	90.3
OmniScience Accuracy	44.3	31.1
OmniScience Non-Hallucination	38.4	74.4
PaperBench	63.7	64.8
scicode	52.1	53.5
simplebench	45.8	70.4
simpleqa_verified	38.9	58.5
SWE-bench Multilingual	72	78.3
SWE-bench Pro	55.6	60.6
SWE-bench Verified	80	80.4
TauBench V3 - Banking	11.1	11.8
Terminal-Bench 2.0	54	69.7
Terminal-Bench Hard	62.2	50.8
τ²-Bench Telecom (AA run)	98.7	94.7
τ³-Bench	12	10.9

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.