Which is better, DeepSeek-V4-Pro or Qwen3.6 Plus?

Across 52 shared benchmarks tracked on Vector Wire, DeepSeek-V4-Pro leads on 40 and Qwen3.6 Plus leads on 12 — DeepSeek-V4-Pro 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 Qwen3.6 Plus

DeepSeekvsAlibaba52 shared benchmarks40–12 head-to-head

Benchmark	DeepSeek-V4-Pro	Qwen3.6 Plus
AA Agentic Index	63.3	29
AA Intelligence	53	40.5
AA-LCR	70	72.3
AA-Omniscience	-10.6	2.6
AgentWorldBench - Android	55.2	57.6
AgentWorldBench - MCP	63.3	55.3
AgentWorldBench - OS	63.7	60.3
AgentWorldBench - Overall	53	50.8
AgentWorldBench - Search	27.6	21.9
AgentWorldBench - SWE	59.4	59.1
AgentWorldBench - Terminal	51.3	50.6
AgentWorldBench - Web	50.3	50.8
aime_2026	96.7	95.3
arena_elo	1457	1444
arena_text_factuality	1450	1438
Artificial Analysis Coding Index	59.4	54.5
coding_arena_elo	1582	1460
critpt	13	2.9
gdpval	49	32
GPQA Diamond	90.5	90.4
HLE	48.2	28.8
HLE (with tools)	48.2	50.6
hmmt_feb_2026	95.2	87.8
hmmt_nov_2025	94.4	94.6
IFBench	76.5	75.2
imo_answer_bench	89.8	83.8
livebench	73.6	70.8
livebench_agentic_coding	42.6	41.4
livebench_coding	70	78.2
livebench_data_analysis	74.5	69.9
livebench_instruction_following	62.4	58.3
livebench_language	78.1	75
livebench_math	90.7	83.7
livebench_reasoning	82.7	75.8
MCP Atlas	74.2	74.1
MMLU-Pro	87.5	88.5
nl2repo	38.5	37.9
OmniScience Accuracy	43	26.4
OmniScience Non-Hallucination	12.2	68
scicode	50	40.7
simpleqa_verified	57.9	49.1
SWE-bench Multilingual	76.2	73.8
SWE-bench Pro	55.4	56.6
SWE-bench Verified	80.6	78.8
TauBench V3 - Banking	30.1	20.8
Terminal-Bench 2.0	67.9	61.6
Terminal-Bench 2.1	72.1	61.4
Terminal-Bench Hard	46.2	43.9
tool_decathlon	52.8	39.8
toolathlon	51.8	39.8
τ²-Bench Telecom (AA run)	96.2	97.7
τ³-Bench	25.8	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.