benchgap
StepFun · model

Step 5 Preview benchmark scores

As of 2026-11-21, Step 5 Preview (StepFun) has measured scores on 39 benchmarks and estimated scores on 139 more.

BenchmarkScoreSource
AA-GPQA Diamond93.2%estimated ± 4.4 pp, high confidence
AA-HLE46.5%measured
AA-SciCode58.9%measured
Artificial Analysis Intelligence Index28.5%measured
CritPt20.9%measured
GDPval-AA54.3%measured
GPQA Diamond93.5%measured
HLE46.5%measured
AA-LCR88.3%measured
LiveCodeBench v693.1%estimated ± 4.4 pp, high confidence
AA-Omniscience Accuracy36.4%measured
SciCode58.9%measured
Terminal-Bench 2.185.0%measured
BrowseComp88.7%measured
HLE w/ tools55.0%estimated ± 3.1 pp, high confidence
LongBench v273.6%estimated ± 5.5 pp, low confidence
VITA-Bench46.9%estimated ± 6.7 pp, low confidence
τ²-bench results95.4%estimated ± 7.0 pp, medium confidence
AA Coding Index75.5%estimated ± 2.5 pp, high confidence
AA-MMMU-Pro76.4%measured
APEX-Agents37.8%measured
APEX-Agents-AA38.0%measured
BioMysteryBench (human-difficult)48.3%estimated ± 6.8 pp, low confidence
DeepSearchQA91.2%estimated ± 6.8 pp, low confidence
SWE-bench Verified87.0%estimated ± 5.1 pp, medium confidence
AutomationBench44.0%measured
BFCL v477.0%estimated ± 2.7 pp, medium confidence
CyberGym84.7%measured
JobBench59.0%measured
SkillsBench61.5%estimated ± 4.1 pp, medium confidence
SWE-bench Pro73.8%estimated ± 6.3 pp, medium confidence
WideResearch83.2%estimated ± 1.2 pp, medium confidence
AA Agentic Index50.0%estimated ± 3.1 pp, high confidence
MMLU-Pro85.5%estimated ± 5.0 pp, medium confidence
FailureSensorIQ67.1%estimated ± 1.7 pp, high confidence
GPQA93.5%estimated ± 0.0 pp, high confidence
MMLU90.5%estimated ± 0.9 pp, medium confidence
Vals GPQA Diamond92.5%estimated ± 2.1 pp, high confidence
Vals LiveCodeBench88.5%estimated ± 0.8 pp, low confidence
Vals MMLU-Pro88.5%estimated ± 1.6 pp, high confidence
MMMLU90.9%estimated ± 0.7 pp, low confidence
MMMU83.8%estimated ± 0.8 pp, high confidence
Chess Puzzles38.8%estimated ± 10.8 pp, low confidence
AA MMLU-Pro86.6%estimated ± 0.8 pp, high confidence
Gert Labs69.7%estimated ± 5.5 pp, low confidence
AA-AnalystAgent53.3%estimated ± 3.7 pp, medium confidence
AA EnterpriseOps-Gym47.2%measured
AA Harvey LAB v1.093.4%measured
AA Tau3 Banking47.1%estimated ± 2.4 pp, low confidence
AA Terminal-Bench 2.183.0%estimated ± 2.6 pp, high confidence
ApprenticeBench26.6%estimated ± 4.0 pp, medium confidence
ARC-AGI-194.2%estimated ± 2.5 pp, high confidence
ARC-AGI-276.1%estimated ± 9.7 pp, medium confidence
CursorBench 3.165.9%estimated ± 3.4 pp, medium confidence
CursorBench 3.265.3%estimated ± 3.6 pp, medium confidence
EBR-bench30.3%estimated ± 11.3 pp, low confidence
Terminal-Bench 3.018.3%estimated ± 9.1 pp, low confidence
FrontierCode 1.1 Main48.6%estimated ± 3.5 pp, low confidence
FrontierSWE v239.2%estimated ± 14.2 pp, low confidence
Furniture Assembly36.2%estimated ± 12.4 pp, low confidence
MaxxBench49.3%estimated ± 5.7 pp, medium confidence
MazeBench2.7%estimated ± 4.0 pp, high confidence
Mystery Game Puzzles25.3%estimated ± 13.3 pp, low confidence
NYT Connections76.6%measured
OfficeQA Pro60.3%measured
OSWorld-Verified80.0%estimated ± 4.7 pp, high confidence
ParseBench-167.9%estimated ± 5.4 pp, low confidence
Terminal-Bench 4.125.3%estimated ± 4.6 pp, high confidence
Vals SWE-bench77.8%estimated ± 5.3 pp, low confidence
Terminal-Bench 2.1 (Vals AI)71.4%estimated ± 7.3 pp, medium confidence
CWE-Bench-AA52.5%measured
AA Cyber Index47.7%measured
DeepsecBench-AA15.8%measured
HealthBench Professional (length-adjusted, Artificial Analysis)54.5%estimated ± 4.3 pp, low confidence
HealthBench Professional (raw, Artificial Analysis)66.0%estimated ± 0.8 pp, medium confidence
AA ITBench55.6%measured
Bug Hunt Bench62.4%estimated ± 3.2 pp, low confidence
Chartography (tools)81.1%estimated ± 4.1 pp, low confidence
CursorBench 4.053.9%estimated ± 4.7 pp, high confidence
CWE-bench v158.6%estimated ± 6.4 pp, medium confidence
DeepSWE67.7%measured
DRACO83.3%measured
FrontierCode 2.026.3%estimated ± 9.3 pp, low confidence
FrontierPhysics1.7%estimated ± 4.7 pp, high confidence
HLE w/o tools39.4%estimated ± 2.7 pp, high confidence
OfficeQA73.9%estimated ± 0.2 pp, low confidence
OSWorld 2.019.4%estimated ± 5.5 pp, low confidence
OSWorld 2.169.5%estimated ± 6.0 pp, medium confidence
OSWorld 2.1 Offline69.6%estimated ± 1.9 pp, medium confidence
PostTrainBench v1.135.6%estimated ± 4.5 pp, high confidence
ProgramBench80.5%measured
SpireClimb5.8%estimated ± 7.2 pp, medium confidence
StealthMark39.0%measured
SWE-bench Multimodal36.7%estimated ± 8.8 pp, low confidence
Terminal-Bench 4.033.3%measured
Terminal-Bench-Science 0.17.2%estimated ± 14.2 pp, low confidence
Toolathlon-Verified74.1%measured
VulcanBench Frontier v492.2%estimated ± 1.2 pp, medium confidence
AA AutomationBench51.7%estimated ± 5.5 pp, medium confidence
GDP.pdf8.6%estimated ± 3.1 pp, medium confidence
AA Terminal-Bench 4.030.8%estimated ± 6.3 pp, low confidence
ARC-AGI-30.0%estimated ± 8.3 pp, low confidence
ARC-AGI-3 Provider Adapter11.6%estimated ± 7.3 pp, medium confidence
cursorBench4145.1%estimated ± 3.7 pp, high confidence
CWE-bench42.0%estimated ± 4.3 pp, low confidence
MKA27.6%measured
SkillsBench official leaderboard (BenchFlow)53.7%estimated ± 6.1 pp, medium confidence
BinaryAudit35.8%estimated ± 12.1 pp, low confidence
OTelBench22.5%estimated ± 6.6 pp, medium confidence
FACTS Multimodal43.3%estimated ± 4.0 pp, high confidence
Hermes Bench69.2%measured
Hermes Index32.8%measured
ParseBench-266.2%estimated ± 5.0 pp, medium confidence
CharXiv87.0%estimated ± 5.1 pp, low confidence
CharXiv w/o tools80.2%estimated ± 1.2 pp, high confidence
ExploitBench26.1%estimated ± 12.6 pp, low confidence
ExploitGym9.2%estimated ± 6.4 pp, low confidence
AI2D_TEST93.1%estimated ± 2.0 pp, medium confidence
C-Eval94.1%estimated ± 1.0 pp, low confidence
Claw-Eval75.6%estimated ± 4.5 pp, medium confidence
CountBench95.9%estimated ± 2.9 pp, medium confidence
DeepPlanning52.6%estimated ± 13.1 pp, low confidence
DynaMath85.4%estimated ± 3.0 pp, medium confidence
ERQA66.5%estimated ± 3.1 pp, high confidence
MathVision85.1%estimated ± 3.0 pp, high confidence
MCP Atlas85.6%measured
MLVU (M-Avg)85.0%estimated ± 2.2 pp, high confidence
MMLU-Redux95.5%estimated ± 0.8 pp, medium confidence
MMMU-Pro76.0%measured
NL2Repo59.8%estimated ± 6.0 pp, medium confidence
QwenClawBench70.7%estimated ± 2.6 pp, medium confidence
RealWorldQA83.2%estimated ± 2.8 pp, high confidence
ScreenSpot Pro74.1%estimated ± 9.6 pp, low confidence
SimpleVQA63.9%estimated ± 7.5 pp, medium confidence
SuperGPQA86.0%estimated ± 7.2 pp, medium confidence
Toolathlon59.2%estimated ± 4.2 pp, high confidence
V*100.0%estimated ± 7.5 pp, medium confidence
Video-MME (w/o subtitle)83.4%estimated ± 1.3 pp, high confidence
Video-MME (with subtitle)85.9%estimated ± 3.0 pp, high confidence
VideoMMMU84.1%estimated ± 2.4 pp, high confidence
We-Math81.4%estimated ± 7.4 pp, low confidence
HealthBench Hard32.7%estimated ± 8.6 pp, low confidence
MedXpertQA (MM)65.6%estimated ± 2.4 pp, high confidence
MedXpertQA (Text)65.9%estimated ± 4.5 pp, medium confidence
MMLU-Pro (Arcee)88.0%estimated ± 3.4 pp, low confidence
React Native Evals80.5%estimated ± 4.7 pp, low confidence
ResearchClawBench19.9%estimated ± 1.8 pp, medium confidence
DAYJOB0.1%estimated ± 3.2 pp, high confidence
DAYJOB Healthcare0.0%estimated ± 4.7 pp, high confidence
BioMysteryBench (human-solvable)89.2%estimated ± 0.6 pp, medium confidence
HealthBench (raw)54.5%estimated ± 6.0 pp, medium confidence
HealthBench (length-adjusted)56.0%estimated ± 3.4 pp, medium confidence
HealthBench Professional57.9%estimated ± 4.2 pp, medium confidence
HealthBench Professional (raw)64.8%estimated ± 4.5 pp, high confidence
HLE-Verified54.3%estimated ± 6.8 pp, low confidence
LABBench282.6%estimated ± 2.4 pp, medium confidence
ScarfBench65.6%estimated ± 3.8 pp, medium confidence
BlenderWorld (text-only)72.5%estimated ± 4.3 pp, medium confidence
Cookie-Bench83.3%estimated ± 4.8 pp, medium confidence
STEMBench2.5%estimated ± 0.2 pp, medium confidence
SimpleQA40.6%estimated ± 4.3 pp, low confidence
Agents' Last Exam29.5%measured
BabyVision w/ Python88.8%estimated ± 3.8 pp, low confidence
OpenHarmony Bench v1.058.9%estimated ± 2.4 pp, high confidence
SEC-Bench Pro45.4%estimated ± 9.3 pp, medium confidence
BabyVision56.1%estimated ± 9.1 pp, low confidence
MMVU76.0%estimated ± 2.7 pp, medium confidence
PerceptionBench58.6%estimated ± 4.9 pp, low confidence
MRCRv286.0%estimated ± 11.3 pp, low confidence
MMLongBench-Doc61.7%estimated ± 2.6 pp, low confidence
OCRBench V270.6%estimated ± 3.3 pp, high confidence
SWE-Marathon72.7%measured
BioBasicsBench36.1%estimated ± 2.8 pp, medium confidence
LifeVal78.8%estimated ± 1.6 pp, medium confidence
MMMU-Pro w/ Python79.5%estimated ± 2.0 pp, high confidence
ExploitBench Internal Port0.0%estimated ± 6.8 pp, low confidence
PinchBench90.6%estimated ± 3.6 pp, low confidence
SWE-Atlas Codebase QnA59.9%estimated ± 4.8 pp, high confidence