Which is better, Gemini 3 Pro or Qwen3.5 35B A3B?

Across 53 shared benchmarks tracked on Vector Wire, Gemini 3 Pro leads on 39 and Qwen3.5 35B A3B leads on 14 — 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 35B A3B

GooglevsAlibaba53 shared benchmarks39–14 head-to-head

Benchmark	Gemini 3 Pro	Qwen3.5 35B A3B
AA Agentic Index	52	44.1
AA Intelligence	40.6	29.9
AA-LCR	73	68.3
AA-Omniscience	15.3	-48.1
aime_2026	91.7	93.3
Artificial Analysis Coding Index	46.5	37
browsecomp	59.2	61
browsecomp_zh	66.8	69.5
charxiv_rq	81.4	77.5
coding_arena_elo	1438	1250
critpt	9.1	0.9
gdpval	34.2	20.3
GPQA Diamond	91.9	84.5
HLE	45.8	47.4
hmmt_feb_2025	97.5	89
hmmt_feb_2026	86.4	81.8
hmmt_nov_2025	93.3	89.2
IFBench	70.4	72.5
imo_answer_bench	83.3	76.8
LiveCodeBench v6	90.7	74.6
longbench_v2	68.2	59
LVBench	73.5	71.4
mathvision	86.1	83.9
MathVista	89.8	86.2
MCP Atlas	70.3	62.4
MLVU	80.7	85.6
MMLU-Pro	90.1	85.3
mmmlu	91.8	85.2
MMMU-Pro	81	75.1
MMVU	78.9	72.3
multichallenge	65.7	60
OCRBench	90.3	91
ocrbench_v2	63.4	65.3
OmniDocBench 1.5	88.5	89.3
OmniScience Accuracy	55.8	20.1
OmniScience Non-Hallucination	10	14.6
scicode	56.1	37.7
screenspot_pro_no_tools	72.7	68.6
Seal-0	45.5	41.4
SimpleVQA	69.7	58.3
SWE-bench Multilingual	68.7	60.3
SWE-bench Pro	43.3	44.6
SWE-bench Verified	78	70
Terminal-Bench 2.0	54.2	40.5
Terminal-Bench Hard	56.9	26.5
tool_decathlon	36.4	28.7
vectara_answer_rate	99.4	99.8
vectara_avg_summary_length	101.9	94.9
vectara_factual_consistency	86.4	89.5
vectara_hallucination_rate ↓	13.6	10.5
VideoMMMU	87.6	80.4
ZeroBench	8	0.1
τ²-Bench Telecom (AA run)	98	89.2

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.