Which is better, Gemma 4 31B or Qwen3.7 Max?

Across 43 shared benchmarks tracked on Vector Wire, Gemma 4 31B leads on 2 and Qwen3.7 Max leads on 41 — 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 —

Gemma 4 31B vs Qwen3.7 Max

GooglevsAlibaba43 shared benchmarks2–41 head-to-head

Benchmark	Gemma 4 31B	Qwen3.7 Max
AA Agentic Index	14.4	30.9
AA Intelligence	29.7	46.7
AA-LCR	68.3	74.7
AA-Omniscience	-47.9	13.5
aime_2026	89.2	97
arena_elo	1451	1474
arena_text_factuality	1449	1479
Artificial Analysis Coding Index	43.4	66
Claw Eval (pass@3)	25	65.2
coding_arena_elo	1364	1517
critpt	1.4	13.4
gdpval	15.5	38.5
GPQA Diamond	85.7	92.4
HLE	26.5	41.4
HLE (with tools)	26.5	53.5
hmmt_feb_2026	77.2	97.1
hmmt_nov_2025	87.5	95
IFBench	75.6	80.5
imo_answer_bench	74.5	90
itbenchSre	37.3	42.5
LiveCodeBench v6	80	91.6
MCP Atlas	57.2	76.4
mcpmark	18.1	60.8
mmlu_redux	93.7	95
MMLU-Pro	85.2	89.6
mmmlu	88.4	90.3
nl2repo	15.5	47.2
OmniScience Accuracy	20	31.1
OmniScience Non-Hallucination	18.1	74.4
scicode	43.4	53.5
simpleqa_verified	9.6	58.5
supergpqa	65.7	73.6
SWE-bench Multilingual	51.7	78.3
SWE-bench Pro	35.7	60.6
SWE-bench Verified	52	80.4
TauBench V3 - Banking	14.8	11.8
Terminal-Bench 2.0	42.9	69.7
Terminal-Bench 2.1	43.4	75
Terminal-Bench Hard	36.4	50.8
VITA-Bench	43	47.9
WideSearch	35.2	75.2
τ²-Bench Telecom (AA run)	65.5	94.7
τ³-Bench	67.5	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.