Which is better, Gemini 3 Pro or Qwen3.5 397B A17B?

Across 46 shared benchmarks tracked on Vector Wire, Gemini 3 Pro leads on 29 and Qwen3.5 397B A17B leads on 17 — Gemini 3 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 Pro vs Qwen3.5 397B A17B

GooglevsAlibaba46 shared benchmarks29–17 head-to-head

Benchmark	Gemini 3 Pro	Qwen3.5 397B A17B
AA Agentic Index	52	53.3
AA Intelligence	40.6	34.3
AA-LCR	73	72.7
AA-Omniscience	15.3	-30.8
aime_2026	91.7	94.2
apexAgents	18.4	15.3
arena_elo	1485	1442
arena_text_factuality	1481	1445
arena_vision	1290	1249
Artificial Analysis Coding Index	46.5	48.2
browsecomp	59.2	69
BrowseComp (context management)	67.6	78.6
browsecomp_zh	66.8	70.3
charxiv_rq	81.4	80.8
coding_arena_elo	1438	1399
critpt	9.1	1.7
gdpval	34.2	35.8
GPQA Diamond	91.9	89.3
HLE	45.8	29
HLE (with tools)	45.8	48.3
hmmt_feb_2025	97.5	94.8
hmmt_feb_2026	86.4	87.9
hmmt_nov_2025	93.3	92.7
IFBench	70.4	78.8
imo_answer_bench	83.3	80.9
LiveCodeBench v6	90.7	83.6
longbench_v2	68.2	63.2
MLVU	80.7	86.7
MMLU-Pro	90.1	87.8
mmmlu	91.8	88.5
MMMU-Pro	81	79
multichallenge	65.7	67.6
OmniScience Accuracy	55.8	30.8
OmniScience Non-Hallucination	10	17.3
scicode	56.1	42
Seal-0	45.5	46.9
simpleqa_verified	72.9	26
SimpleVQA	69.7	67.1
SWE-bench Multilingual	68.7	69.3
SWE-bench Pro	43.3	50.9
SWE-bench Verified	78	76.4
Terminal-Bench 2.0	54.2	52.5
Terminal-Bench Hard	56.9	40.9
toolathlon	36.4	38.3
VideoMMMU	87.6	84.7
τ²-Bench Telecom (AA run)	98	95.6

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.