Which is better, Qwen3.5 35B A3B or Qwen3.5 397B A17B?

Across 75 shared benchmarks tracked on Vector Wire, Qwen3.5 35B A3B leads on 5 and Qwen3.5 397B A17B leads on 70 — Qwen3.5 397B A17B leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Qwen3.5 35B A3B vs Qwen3.5 397B A17B

AlibabavsAlibaba75 shared benchmarks5–70 head-to-head

Benchmark	Qwen3.5 35B A3B	Qwen3.5 397B A17B
AA Agentic Index	44.1	53.3
AA Intelligence	29.9	34.3
AA-LCR	68.3	72.7
AA-Omniscience	-48.1	-30.8
AgentWorldBench - Android	53.2	54.9
AgentWorldBench - MCP	57.9	68.3
AgentWorldBench - OS	56.3	60.9
AgentWorldBench - Overall	47.7	54.7
AgentWorldBench - Search	26	30.8
AgentWorldBench - SWE	47.6	64.4
AgentWorldBench - Terminal	46.1	55.3
AgentWorldBench - Web	47.1	48.5
aime_2026	93.3	94.2
Artificial Analysis Coding Index	37	48.2
bfcl_v4	67.3	72.9
browsecomp	61	69
browsecomp_zh	69.5	70.3
C-Eval	90.2	93
CC-OCR	80.7	82
charxiv_rq	77.5	80.8
Claw Eval (pass@3)	51	48.1
Claw-Eval Avg	65.4	70.7
coding_arena_elo	1250	1399
critpt	0.9	1.7
DeepPlanning	22.8	34.3
ERQA	64.8	67.5
gdpval	20.3	35.8
GPQA Diamond	84.5	89.3
GSM8K	90.1	66.7
HLE	47.4	29
HMMT 2025	89.2	92.7
hmmt_feb_2025	89	94.8
hmmt_feb_2026	81.8	87.9
hmmt_nov_2025	89.2	92.7
IFBench	72.5	78.8
ifeval	91.9	92.6
imo_answer_bench	76.8	80.9
include	79.7	85.6
itbenchSre	21.5	34.1
LiveCodeBench v6	74.6	83.6
longbench_v2	59	63.2
mcpmark	27	46.1
MLVU	85.6	86.7
mmlu_prox	81	84.7
mmlu_redux	93.3	94.9
MMLU-Pro	85.3	87.8
mmmlu	85.2	88.5
MMMU	81.4	85
MMMU-Pro	75.1	79
MMStar	81.9	83.8
multichallenge	60	67.6
MV-Bench	74.8	77.6
nl2repo	20.5	32.2
OmniScience Accuracy	20.1	30.8
OmniScience Non-Hallucination	14.6	17.3
QwenClawBench	47.7	51.8
QwenWebBench	978	1186
RealWorldQA	84.1	83.9
scicode	37.7	42
Seal-0	41.4	46.9
SimpleVQA	58.3	67.1
SkillsBench Avg5	4.4	30
supergpqa	63.4	70.4
SWE-bench Multilingual	60.3	69.3
SWE-bench Pro	44.6	50.9
SWE-bench Verified	70	76.4
TauBench V3 - Banking	4.9	13.4
Terminal-Bench 2.0	40.5	52.5
Terminal-Bench 2.1	40.8	51.3
Terminal-Bench Hard	26.5	40.9
VideoMMMU	80.4	84.7
VITA-Bench	31.9	49.7
WideSearch	59.1	74
τ²-Bench Telecom (AA run)	89.2	95.6
τ³-Bench	68.9	13.4

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.