benchgap
Alibaba (Qwen) · model

Qwen4 Max benchmark scores

As of 2026-11-18, Qwen4 Max (Alibaba (Qwen)) has measured scores on 15 benchmarks and estimated scores on 76 more.

BenchmarkScoreSource
AA-GPQA Diamond92.8%estimated ± 2.1 pp, medium confidence
AA-HLE50.5%estimated ± 4.6 pp, high confidence
AA-SciCode59.0%estimated ± 3.9 pp, medium confidence
Artificial Analysis Intelligence Index42.9%measured
CritPt21.3%estimated ± 5.3 pp, medium confidence
GDPval-AA61.7%estimated ± 8.6 pp, medium confidence
GPQA Diamond94.1%estimated ± 1.2 pp, medium confidence
HLE51.7%measured
AA-Omniscience Accuracy43.7%measured
SciCode62.2%estimated ± 7.6 pp, low confidence
Terminal-Bench 2.189.1%estimated ± 5.4 pp, low confidence
BrowseComp93.0%estimated ± 4.2 pp, medium confidence
HLE w/ tools65.5%estimated ± 6.8 pp, low confidence
τ²-bench results95.7%estimated ± 7.0 pp, low confidence
AA Coding Index77.9%estimated ± 2.5 pp, medium confidence
APEX-Agents-AA59.1%measured
BioMysteryBench (human-difficult)44.4%estimated ± 7.8 pp, low confidence
DeepSearchQA100.0%estimated ± 7.9 pp, low confidence
SWE-bench Verified90.8%estimated ± 5.4 pp, low confidence
AutomationBench47.3%estimated ± 6.1 pp, low confidence
JobBench81.7%estimated ± 8.2 pp, low confidence
SWE-bench Pro92.2%estimated ± 7.7 pp, low confidence
AA Agentic Index49.1%estimated ± 12.2 pp, low confidence
MMLU-Pro85.9%estimated ± 2.5 pp, low confidence
GPQA94.1%estimated ± 1.2 pp, medium confidence
MMLU90.5%estimated ± 0.9 pp, medium confidence
Vals GPQA Diamond93.5%estimated ± 1.2 pp, medium confidence
Vals LiveCodeBench77.4%estimated ± 6.1 pp, low confidence
Vals MMLU-Pro90.2%estimated ± 1.8 pp, high confidence
MMMLU94.6%estimated ± 3.0 pp, low confidence
AA MMLU-Pro87.7%estimated ± 0.8 pp, high confidence
Gert Labs74.8%estimated ± 10.4 pp, low confidence
AA-AnalystAgent46.3%measured
AA EnterpriseOps-Gym48.5%measured
ARC-AGI-196.2%estimated ± 3.9 pp, high confidence
ARC-AGI-281.7%estimated ± 13.0 pp, low confidence
CursorBench 3.231.9%estimated ± 3.6 pp, low confidence
FrontierCode 1.1 Main52.8%estimated ± 3.5 pp, low confidence
MazeBench3.4%estimated ± 6.2 pp, low confidence
Terminal-Bench 4.148.3%estimated ± 5.0 pp, medium confidence
Vals SWE-bench85.5%estimated ± 10.6 pp, low confidence
Terminal-Bench 2.1 (Vals AI)78.5%estimated ± 6.9 pp, low confidence
CWE-Bench-AA64.1%estimated ± 9.0 pp, medium confidence
DeepsecBench-AA39.4%estimated ± 7.5 pp, medium confidence
Harvey LAB-AA v1.17.5%estimated ± 3.5 pp, medium confidence
HealthBench Professional (length-adjusted, Artificial Analysis)58.9%estimated ± 4.3 pp, medium confidence
HealthBench Professional (raw, Artificial Analysis)67.2%estimated ± 1.1 pp, medium confidence
AA ITBench47.8%estimated ± 6.3 pp, low confidence
Bug Hunt Bench38.2%estimated ± 10.4 pp, low confidence
Chartography (no tools)54.2%measured
Chartography (tools)87.4%estimated ± 2.3 pp, medium confidence
CursorBench 4.048.0%estimated ± 4.7 pp, high confidence
DeepSWE75.7%measured
HLE w/o tools51.7%measured
OSWorld 2.1 Offline59.6%estimated ± 7.9 pp, low confidence
PostTrainBench v1.144.2%estimated ± 4.5 pp, high confidence
ProgramBench31.5%measured
SWE-bench Multimodal56.1%estimated ± 8.3 pp, low confidence
Terminal-Bench 4.048.5%measured
Terminal-Bench-Science 0.137.1%estimated ± 14.9 pp, low confidence
Toolathlon-Verified76.6%estimated ± 2.8 pp, low confidence
VulcanBench Frontier v488.5%estimated ± 1.2 pp, medium confidence
AA AutomationBench72.6%measured
GDP.pdf26.2%measured
AA Terminal-Bench 4.048.5%measured
ARC-AGI-313.6%estimated ± 7.2 pp, medium confidence
ARC-AGI-3 Provider Adapter8.1%estimated ± 11.0 pp, low confidence
CWE-bench54.5%estimated ± 6.2 pp, low confidence
MKA49.3%estimated ± 5.3 pp, medium confidence
Hermes Bench76.4%estimated ± 1.8 pp, medium confidence
Hermes Index63.3%estimated ± 2.7 pp, medium confidence
ExploitGym33.1%estimated ± 9.8 pp, low confidence
C-Eval98.4%estimated ± 1.0 pp, low confidence
MCP Atlas97.8%estimated ± 9.1 pp, low confidence
MMLU-Redux96.4%estimated ± 0.8 pp, low confidence
NL2Repo63.3%estimated ± 6.0 pp, medium confidence
SuperGPQA90.8%estimated ± 7.2 pp, medium confidence
Toolathlon64.7%estimated ± 6.7 pp, low confidence
MedXpertQA (Text)68.5%estimated ± 8.9 pp, low confidence
ResearchClawBench20.3%estimated ± 2.4 pp, low confidence
HealthBench (length-adjusted)53.3%estimated ± 2.2 pp, high confidence
HealthBench Professional60.0%measured
HLE-Verified38.8%estimated ± 7.1 pp, low confidence
ScarfBench69.7%estimated ± 3.8 pp, medium confidence
STEMBench2.5%estimated ± 0.2 pp, medium confidence
SimpleQA50.7%estimated ± 4.3 pp, low confidence
Agents' Last Exam47.0%estimated ± 12.9 pp, low confidence
SWE-Marathon46.1%estimated ± 12.9 pp, low confidence
BioBasicsBench34.1%estimated ± 3.8 pp, medium confidence
LifeVal77.8%estimated ± 2.3 pp, low confidence
SWE-Atlas Codebase QnA61.7%estimated ± 4.8 pp, medium confidence