benchgap
Other · model

Hy4 flash benchmark scores

As of 2026-11-18, Hy4 flash (Other) has measured scores on 9 benchmarks and estimated scores on 74 more.

BenchmarkScoreSource
AA-GPQA Diamond89.1%estimated ± 7.1 pp, medium confidence
AA-HLE44.8%estimated ± 6.5 pp, medium confidence
AA-SciCode56.4%estimated ± 3.9 pp, medium confidence
Artificial Analysis Intelligence Index37.7%estimated ± 10.7 pp, low confidence
CritPt18.4%estimated ± 6.1 pp, medium confidence
GDPval-AA56.5%estimated ± 4.2 pp, high confidence
GPQA Diamond92.6%estimated ± 2.3 pp, high confidence
HLE55.2%measured
AA-Omniscience Accuracy47.1%estimated ± 9.0 pp, medium confidence
SciCode57.0%estimated ± 7.6 pp, low confidence
BrowseComp89.9%estimated ± 4.8 pp, medium confidence
HLE w/ tools55.2%measured
τ²-bench results97.8%estimated ± 3.1 pp, high confidence
AA Coding Index74.5%estimated ± 2.5 pp, high confidence
AA-MMMU-Pro75.0%estimated ± 1.5 pp, high confidence
APEX-Agents63.1%measured
SWE-bench Verified85.6%estimated ± 5.4 pp, medium confidence
AutomationBench52.2%measured
CyberGym87.6%estimated ± 4.6 pp, medium confidence
SWE-bench Pro65.5%estimated ± 7.7 pp, medium confidence
WideResearch80.2%estimated ± 4.8 pp, medium confidence
AA Agentic Index40.2%estimated ± 9.2 pp, medium confidence
GPQA92.1%estimated ± 1.9 pp, high confidence
Vals LiveCodeBench77.4%estimated ± 6.1 pp, low confidence
Vals MMLU-Pro88.6%estimated ± 2.4 pp, medium confidence
MMMU83.1%estimated ± 1.7 pp, high confidence
AA Harvey LAB v1.089.5%estimated ± 1.7 pp, medium confidence
AA Tau3 Banking50.2%estimated ± 3.3 pp, low confidence
ARC-AGI-192.7%estimated ± 11.6 pp, low confidence
CursorBench 3.217.6%estimated ± 3.6 pp, low confidence
FrontierCode 1.1 Main46.9%estimated ± 3.5 pp, low confidence
ParseBench-158.1%estimated ± 8.4 pp, medium confidence
Terminal-Bench 4.116.0%estimated ± 5.0 pp, medium confidence
Vals SWE-bench85.5%estimated ± 10.6 pp, low confidence
CWE-Bench-AA56.2%estimated ± 5.2 pp, medium confidence
AA Cyber Index46.3%estimated ± 9.4 pp, medium confidence
Bug Hunt Bench26.4%estimated ± 10.4 pp, low confidence
CursorBench 4.040.8%estimated ± 4.7 pp, medium confidence
DeepSWE64.6%measured
HLE w/o tools46.1%estimated ± 6.0 pp, medium confidence
OSWorld 2.168.1%measured
PostTrainBench v1.132.7%estimated ± 4.5 pp, high confidence
ProgramBench16.0%measured
Terminal-Bench 4.019.2%measured
Terminal-Bench-Science 0.10.3%estimated ± 14.9 pp, low confidence
Toolathlon-Verified70.3%estimated ± 5.7 pp, medium confidence
VulcanBench Frontier v488.5%estimated ± 1.2 pp, low confidence
AA AutomationBench41.9%estimated ± 7.9 pp, low confidence
GDP.pdf14.7%estimated ± 6.1 pp, low confidence
AA Terminal-Bench 4.018.0%estimated ± 6.3 pp, low confidence
ARC-AGI-34.8%estimated ± 8.3 pp, medium confidence
CWE-bench62.4%estimated ± 11.0 pp, low confidence
Hermes Bench70.3%estimated ± 1.8 pp, medium confidence
Hermes Index38.2%estimated ± 9.3 pp, low confidence
ParseBench-265.9%estimated ± 5.0 pp, medium confidence
CharXiv76.5%estimated ± 5.1 pp, medium confidence
CharXiv w/o tools79.5%estimated ± 2.1 pp, high confidence
AI2D_TEST91.2%estimated ± 2.0 pp, medium confidence
CountBench94.9%estimated ± 3.0 pp, high confidence
DynaMath84.5%estimated ± 5.3 pp, low confidence
ERQA55.9%estimated ± 6.4 pp, medium confidence
MathVision84.6%estimated ± 2.9 pp, high confidence
MCP Atlas84.7%estimated ± 6.6 pp, low confidence
MLVU (M-Avg)84.9%estimated ± 2.2 pp, high confidence
MMLU-Redux96.9%estimated ± 2.1 pp, medium confidence
MMMU-Pro75.8%measured
NL2Repo58.5%estimated ± 6.0 pp, medium confidence
RealWorldQA83.5%estimated ± 2.9 pp, high confidence
ScreenSpot Pro74.3%estimated ± 10.1 pp, low confidence
SimpleVQA63.7%estimated ± 7.5 pp, medium confidence
SuperGPQA94.1%estimated ± 7.2 pp, low confidence
Video-MME (w/o subtitle)83.3%estimated ± 1.4 pp, medium confidence
Video-MME (with subtitle)85.8%estimated ± 3.0 pp, high confidence
VideoMMMU81.6%estimated ± 5.1 pp, medium confidence
We-Math81.3%estimated ± 7.4 pp, low confidence
MedXpertQA (MM)65.1%estimated ± 2.4 pp, high confidence
ScarfBench64.4%estimated ± 3.8 pp, medium confidence
MMVU75.8%estimated ± 2.7 pp, medium confidence
PerceptionBench43.9%estimated ± 2.6 pp, low confidence
SWE-Marathon34.4%estimated ± 12.9 pp, low confidence
MMMU-Pro w/ Python79.2%estimated ± 2.6 pp, high confidence
OCRBench V271.9%estimated ± 4.4 pp, medium confidence
SWE-Atlas Codebase QnA58.8%estimated ± 4.8 pp, high confidence