Which is better, Qwen3.5 397B A17B or Qwen3.7 Max?

Across 47 shared benchmarks tracked on Vector Wire, Qwen3.5 397B A17B leads on 5 and Qwen3.7 Max leads on 42 — Qwen3.7 Max 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 397B A17B vs Qwen3.7 Max

AlibabavsAlibaba47 shared benchmarks5–42 head-to-head

Benchmark	Qwen3.5 397B A17B	Qwen3.7 Max
AA Agentic Index	53.3	30.9
AA Intelligence	34.3	46.7
AA-LCR	72.7	74.7
AA-Omniscience	-30.8	13.5
aime_2026	94.2	97
arena_elo	1442	1474
arena_text_factuality	1445	1479
Artificial Analysis Coding Index	48.2	66
bfcl_v4	72.9	75
Claw Eval (pass@3)	48.1	65.2
coding_arena_elo	1399	1517
critpt	1.7	13.4
gdpval	35.8	38.5
GPQA Diamond	89.3	92.4
HLE	29	41.4
HLE (with tools)	48.3	53.5
hmmt_feb_2026	87.9	97.1
hmmt_nov_2025	92.7	95
IFBench	78.8	80.5
ifeval	92.6	94.3
imo_answer_bench	80.9	90
include	85.6	86.2
itbenchSre	34.1	42.5
LiveCodeBench v6	83.6	91.6
longbench_v2	63.2	65.3
mcpmark	46.1	60.8
mmlu_prox	84.7	87
mmlu_redux	94.9	95
MMLU-Pro	87.8	89.6
mmmlu	88.5	90.3
nl2repo	32.2	47.2
OmniScience Accuracy	30.8	31.1
OmniScience Non-Hallucination	17.3	74.4
scicode	42	53.5
simpleqa_verified	26	58.5
supergpqa	70.4	73.6
SWE-bench Multilingual	69.3	78.3
SWE-bench Pro	50.9	60.6
SWE-bench Verified	76.4	80.4
TauBench V3 - Banking	13.4	11.8
Terminal-Bench 2.0	52.5	69.7
Terminal-Bench 2.1	51.3	75
Terminal-Bench Hard	40.9	50.8
VITA-Bench	49.7	47.9
WideSearch	74	75.2
τ²-Bench Telecom (AA run)	95.6	94.7
τ³-Bench	13.4	10.9

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.