Which is better, Gemma 4 26B A4B or Qwen3.6 27B?

Across 50 shared benchmarks tracked on Vector Wire, Gemma 4 26B A4B leads on 4 and Qwen3.6 27B leads on 46 — Qwen3.6 27B 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 26B A4B vs Qwen3.6 27B

GooglevsAlibaba50 shared benchmarks4–46 head-to-head

Benchmark	Gemma 4 26B A4B	Qwen3.6 27B
AA Agentic Index	32.1	27.5
AA Intelligence	26.1	37.7
AA-LCR	61.7	73.3
AA-Omniscience	-50.8	-20
aime_2026	88.3	94.1
Artificial Analysis Coding Index	39.3	53.7
C-Eval	82.5	91.4
CC-OCR	74.5	81.2
charxiv_rq	69	78.4
Claw Eval (pass@3)	28	60.6
Claw-Eval Avg	58.8	72.4
critpt	0	1.1
gdpval	25.7	32
GPQA Diamond	82.3	87.8
hellaswag	85.3	90
HLE	19.3	24
hmmt_feb_2025	91.7	93.8
hmmt_feb_2026	79	84.3
hmmt_nov_2025	87.5	90.7
humaneval	50	78.2
IFBench	72.4	69.1
imo_answer_bench	74.3	80.8
LiveCodeBench v6	77.1	83.9
mathvision	82.4	85.1
MathVista	79.4	87.4
MMBench EN-DEV-v1.1	89	92.3
mmlu_redux	92.7	93.5
MMLU-Pro	85.2	86.2
MMMU	78.4	82.9
MMMU-Pro	73.8	75.8
nl2repo	11.6	36.2
OmniDocBench 1.5	74.4	89.4
OmniScience Accuracy	19.1	19.6
OmniScience Non-Hallucination	13.6	50.7
QwenClawBench	38.7	53.4
QwenWebBench	1178	1487
RealWorldQA	72.2	84.1
scicode	40.3	39.8
SimpleVQA	52.2	56.1
SkillsBench Avg5	12.3	48.2
supergpqa	61.4	66
SWE-bench Multilingual	43.4	71.3
SWE-bench Pro	13.8	53.5
SWE-bench Verified	57.4	77.2
TauBench V3 - Banking	12	16.7
Terminal-Bench 2.0	34.2	59.3
Terminal-Bench 2.1	39	60.7
Terminal-Bench Hard	25	34.8
τ²-Bench Telecom (AA run)	43.6	94.2
τ³-Bench	59	15.3

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.