benchgap
Alibaba (Qwen) · model

Qwen3.8 27B (xhigh) benchmark scores

As of 2026-10-07, Qwen3.8 27B (xhigh) (Alibaba (Qwen)) has measured scores on 12 benchmarks and estimated scores on 2 more.

BenchmarkScoreSource
Terminal-Bench 2.179.8%measured
Terminal-Bench 4.05.6%measured
Terminal-Bench Science 0.10.5%measured
GPQA Diamond90.5%measured
SciCode46.6%measured
CritPt5.4%measured
GDP.pdf16.6%measured
τ³-Bench Banking48.0%measured
MMMU-Pro76.0%measured
AA-AnalystAgent34.5%estimated ± 3.6 pp, medium confidence
Harvey LAB93.5%estimated ± 1.2 pp, medium confidence
Humanity's Last Exam33.9%measured
AutomationBench48.2%measured
EnterpriseOps-Gym44.2%measured