Which is better, DeepSeek-V4-Pro or Qwen3.7 Max?

Across 45 shared benchmarks tracked on Vector Wire, DeepSeek-V4-Pro leads on 12 and Qwen3.7 Max leads on 33 — 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 —

DeepSeek-V4-Pro vs Qwen3.7 Max

DeepSeekvsAlibaba45 shared benchmarks12–33 head-to-head

Benchmark	DeepSeek-V4-Pro	Qwen3.7 Max
AA Agentic Index	63.3	30.9
AA Intelligence	53	46.7
AA-LCR	70	74.7
AA-Omniscience	-10.6	13.5
aime_2026	96.7	97
arena_elo	1457	1474
arena_text_factuality	1450	1479
Artificial Analysis Coding Index	59.4	66
coding_arena_elo	1582	1517
critpt	13	13.4
DeepSWE	12.8	18
gdpval	49	38.5
GPQA Diamond	90.5	92.4
HLE	48.2	41.4
HLE (with tools)	48.2	53.5
hmmt_feb_2026	95.2	97.1
hmmt_nov_2025	94.4	95
IFBench	76.5	80.5
imo_answer_bench	89.8	90
itbenchSre	38.3	42.5
livebench	73.6	74.3
livebench_agentic_coding	42.6	43.6
livebench_coding	70	74.2
livebench_data_analysis	74.5	71.8
livebench_instruction_following	62.4	74
livebench_language	78.1	79.7
livebench_math	90.7	85.3
livebench_reasoning	82.7	83.3
MCP Atlas	74.2	76.4
MMLU-Pro	87.5	89.6
nl2repo	38.5	47.2
OmniScience Accuracy	43	31.1
OmniScience Non-Hallucination	12.2	74.4
scicode	50	53.5
simplebench	50.9	70.4
simpleqa_verified	57.9	58.5
SWE-bench Multilingual	76.2	78.3
SWE-bench Pro	55.4	60.6
SWE-bench Verified	80.6	80.4
TauBench V3 - Banking	30.1	11.8
Terminal-Bench 2.0	67.9	69.7
Terminal-Bench 2.1	72.1	75
Terminal-Bench Hard	46.2	50.8
τ²-Bench Telecom (AA run)	96.2	94.7
τ³-Bench	25.8	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.