Which is better, Gemini 3.1 Pro or Qwen3.6 Plus?

Across 60 shared benchmarks tracked on Vector Wire, Gemini 3.1 Pro leads on 51 and Qwen3.6 Plus leads on 9 — Gemini 3.1 Pro leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 3.1 Pro vs Qwen3.6 Plus

GooglevsAlibaba60 shared benchmarks51–9 head-to-head

Benchmark	Gemini 3.1 Pro	Qwen3.6 Plus
AA Agentic Index	23	29
AA Intelligence	47.7	40.5
AA-LCR	79	72.3
AA-Omniscience	32.9	2.6
AgentWorldBench - Android	61.4	57.6
AgentWorldBench - MCP	59.1	55.3
AgentWorldBench - OS	66.9	60.3
AgentWorldBench - Overall	54.6	50.8
AgentWorldBench - Search	30.2	21.9
AgentWorldBench - SWE	59.1	59.1
AgentWorldBench - Terminal	52.5	50.6
AgentWorldBench - Web	52.8	50.8
aime_2026	98.3	95.3
arena_elo	1486	1444
arena_text_factuality	1471	1438
Artificial Analysis Coding Index	68.8	54.5
coding_arena_elo	1447	1460
critpt	17.7	2.9
Finance Agent v2	43	40.9
frontiermath_tier_4	16.7	8.3
gdpval	23.3	32
GPQA Diamond	94.3	90.4
HLE	51.4	28.8
HLE (with tools)	51.6	50.6
hmmt_feb_2026	94.7	87.8
hmmt_nov_2025	94.8	94.6
IFBench	77.1	75.2
imo_answer_bench	91	83.8
livebench	79.9	70.8
livebench_agentic_coding	44.1	41.4
livebench_coding	76.5	78.2
livebench_data_analysis	78.5	69.9
livebench_instruction_following	79.1	58.3
livebench_language	85.4	75
livebench_math	91	83.7
livebench_reasoning	84	75.8
LiveCodeBench v6	91.7	87.1
mathvision	89.8	88
MCP Atlas	78.2	74.1
mcpmark	55.9	48.2
MMLU-Pro	91	88.5
mmmlu	92.6	89.5
MMMU-Pro	83	78.8
nl2repo	33.4	37.9
OmniScience Accuracy	55.3	26.4
OmniScience Non-Hallucination	50.1	68
OSWorld-Verified	76.2	62.5
scicode	59	40.7
simpleqa_verified	77.3	49.1
SWE-bench Multilingual	76.9	73.8
SWE-bench Pro	54.2	56.6
SWE-bench Verified	80.6	78.8
TauBench V3 - Banking	21.4	20.8
Terminal-Bench 2.0	68.5	61.6
Terminal-Bench 2.1	74	61.4
Terminal-Bench Hard	68.5	43.9
tool_decathlon	48.8	39.8
toolathlon	48.8	39.8
τ²-Bench Telecom (AA run)	99.3	97.7
τ³-Bench	67.1	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.