benchgap
StepFun · model

Step 3.7 Flash benchmark scores

As of 2026-11-22, Step 3.7 Flash (StepFun) has measured scores on 24 benchmarks and estimated scores on 140 more.

BenchmarkScoreSource
AA-GPQA Diamond80.9%measured
AA-HLE21.4%measured
AA-SciCode43.9%measured
Artificial Analysis Intelligence Index19.5%measured
CritPt2.3%measured
GDPval-AA25.8%measured
GPQA Diamond82.3%estimated ± 2.4 pp, high confidence
HLE24.3%estimated ± 8.5 pp, medium confidence
IFBench65.5%estimated ± 6.6 pp, medium confidence
AA-LCR73.7%measured
LiveCodeBench v688.1%estimated ± 2.7 pp, high confidence
AA-Omniscience Accuracy25.8%measured
SciCode43.9%estimated ± 3.1 pp, high confidence
Terminal-Bench 2.159.5%measured
BrowseComp75.8%measured
HLE w/ tools47.2%measured
IFEval92.7%estimated ± 2.8 pp, high confidence
LongBench v260.8%estimated ± 5.5 pp, medium confidence
VITA-Bench43.3%estimated ± 6.7 pp, low confidence
τ²-bench results98.5%measured
AA Coding Index39.6%measured
AA-MMMU-Pro75.3%measured
APEX-Agents19.3%estimated ± 4.9 pp, low confidence
APEX-Agents-AA14.8%measured
BioMysteryBench (human-difficult)27.5%estimated ± 6.8 pp, low confidence
DeepSearchQA92.8%measured
SWE-bench Verified77.8%estimated ± 3.6 pp, high confidence
BFCL v473.6%estimated ± 2.7 pp, high confidence
CyberGym53.7%estimated ± 6.0 pp, medium confidence
JobBench36.0%estimated ± 7.3 pp, low confidence
SkillsBench47.0%estimated ± 4.1 pp, medium confidence
SWE-bench Pro56.3%measured
WideResearch62.1%estimated ± 3.3 pp, medium confidence
AA Agentic Index19.9%estimated ± 3.1 pp, high confidence
MMLU-Pro85.1%estimated ± 5.0 pp, medium confidence
AA-IFBench67.3%measured
FailureSensorIQ64.3%estimated ± 1.7 pp, high confidence
GPQA82.2%estimated ± 6.1 pp, medium confidence
MMLU90.5%estimated ± 0.9 pp, high confidence
Vals GPQA Diamond80.0%estimated ± 3.2 pp, high confidence
Vals LiveCodeBench82.7%estimated ± 4.7 pp, medium confidence
Vals MMLU-Pro83.1%estimated ± 2.9 pp, high confidence
MMMLU87.3%estimated ± 0.7 pp, medium confidence
MMMU83.6%estimated ± 2.7 pp, high confidence
Chess Puzzles19.6%estimated ± 12.3 pp, low confidence
AA MMLU-Pro85.0%estimated ± 0.8 pp, high confidence
Gert Labs20.3%measured
AA EnterpriseOps-Gym45.8%estimated ± 2.3 pp, medium confidence
AA Harvey LAB v1.086.6%estimated ± 1.7 pp, medium confidence
AA Tau3 Banking39.6%estimated ± 2.0 pp, medium confidence
AA Terminal-Bench 2.181.7%estimated ± 2.6 pp, medium confidence
ApprenticeBench9.3%estimated ± 5.8 pp, low confidence
ARC-AGI-176.5%estimated ± 7.8 pp, low confidence
ARC-AGI-235.0%estimated ± 14.4 pp, low confidence
CursorBench 3.129.0%estimated ± 3.4 pp, low confidence
CursorBench 3.252.6%estimated ± 3.6 pp, high confidence
Terminal-Bench 3.05.2%estimated ± 9.1 pp, low confidence
FrontierCode 1.1 Main13.8%estimated ± 4.6 pp, medium confidence
FrontierSWE v29.2%estimated ± 8.3 pp, medium confidence
MazeBench0.0%estimated ± 2.7 pp, medium confidence
Mystery Game Puzzles13.8%estimated ± 14.6 pp, low confidence
NYT Connections39.7%measured
OfficeQA Pro50.3%estimated ± 6.8 pp, low confidence
OSWorld-Verified72.0%estimated ± 3.8 pp, medium confidence
ParseBench-159.5%estimated ± 6.2 pp, medium confidence
Terminal-Bench 4.10.0%estimated ± 10.9 pp, low confidence
Vals SWE-bench62.5%estimated ± 5.5 pp, medium confidence
Terminal-Bench 2.1 (Vals AI)52.1%estimated ± 7.3 pp, medium confidence
CWE-Bench-AA46.9%estimated ± 5.2 pp, medium confidence
AA Cyber Index3.5%estimated ± 12.4 pp, low confidence
DeepsecBench-AA5.9%estimated ± 8.2 pp, low confidence
HealthBench Professional (length-adjusted, Artificial Analysis)51.7%estimated ± 4.3 pp, low confidence
HealthBench Professional (raw, Artificial Analysis)54.9%estimated ± 1.1 pp, low confidence
AA ITBench43.9%estimated ± 4.1 pp, high confidence
Bug Hunt Bench17.2%estimated ± 10.5 pp, low confidence
Chartography (tools)80.4%estimated ± 4.1 pp, low confidence
CursorBench 4.032.8%estimated ± 3.3 pp, medium confidence
CWE-bench v133.1%estimated ± 5.7 pp, low confidence
DeepSWE43.0%estimated ± 7.5 pp, low confidence
DRACO73.1%estimated ± 2.0 pp, medium confidence
FrontierCode 2.020.1%estimated ± 6.7 pp, low confidence
HLE w/o tools31.4%estimated ± 4.9 pp, high confidence
OSWorld 2.11.3%estimated ± 3.6 pp, medium confidence
OSWorld 2.1 Offline2.9%estimated ± 1.9 pp, low confidence
PostTrainBench v1.123.5%estimated ± 4.4 pp, medium confidence
SpireClimb5.8%estimated ± 7.2 pp, low confidence
StealthMark41.3%estimated ± 10.7 pp, low confidence
Toolathlon-Verified54.9%estimated ± 3.0 pp, high confidence
VulcanBench Frontier v474.0%estimated ± 6.8 pp, low confidence
AA AutomationBench41.9%estimated ± 7.9 pp, low confidence
GDP.pdf9.7%estimated ± 4.2 pp, medium confidence
AA Terminal-Bench 4.04.3%estimated ± 11.0 pp, low confidence
ARC-AGI-31.3%estimated ± 9.8 pp, medium confidence
ARC-AGI-3 Provider Adapter2.7%estimated ± 7.3 pp, low confidence
cursorBench4144.8%estimated ± 3.7 pp, medium confidence
CWE-bench25.8%estimated ± 6.2 pp, low confidence
MKA32.8%estimated ± 4.6 pp, medium confidence
SkillsBench official leaderboard (BenchFlow)44.8%estimated ± 6.1 pp, medium confidence
BinaryAudit35.8%estimated ± 12.1 pp, low confidence
OTelBench22.4%estimated ± 6.6 pp, medium confidence
Vibe Code Bench14.4%estimated ± 10.0 pp, low confidence
FACTS Multimodal23.9%estimated ± 4.0 pp, medium confidence
Hermes Bench66.8%estimated ± 1.8 pp, medium confidence
Hermes Index14.5%estimated ± 5.7 pp, low confidence
ParseBench-265.0%estimated ± 6.2 pp, medium confidence
CharXiv80.0%estimated ± 3.6 pp, high confidence
CharXiv w/o tools80.4%estimated ± 2.5 pp, high confidence
ExploitGym0.0%estimated ± 8.3 pp, low confidence
AI2D_TEST91.9%estimated ± 2.0 pp, medium confidence
C-Eval91.4%estimated ± 1.0 pp, medium confidence
Claw-Eval67.1%measured
CountBench97.4%estimated ± 1.6 pp, medium confidence
DynaMath85.3%estimated ± 3.0 pp, medium confidence
ERQA64.2%estimated ± 3.1 pp, high confidence
MathVision85.6%estimated ± 3.1 pp, high confidence
MCP Atlas42.0%estimated ± 9.1 pp, medium confidence
MCP-Tasks76.3%estimated ± 7.0 pp, low confidence
MLVU (M-Avg)88.2%estimated ± 5.0 pp, high confidence
MMLU-Redux94.2%estimated ± 0.8 pp, medium confidence
MMMU-Pro76.5%estimated ± 1.5 pp, high confidence
NL2Repo27.3%estimated ± 4.7 pp, medium confidence
OmniDocBench 1.591.1%estimated ± 2.0 pp, medium confidence
PostTrainBench v121.4%estimated ± 4.6 pp, high confidence
QwenClawBench55.9%estimated ± 6.0 pp, low confidence
RealWorldQA83.2%estimated ± 2.9 pp, medium confidence
ScreenSpot Pro65.9%estimated ± 7.5 pp, low confidence
SimpleVQA79.2%measured
SuperGPQA66.0%estimated ± 9.3 pp, medium confidence
Toolathlon49.5%measured
V*95.3%measured
Video-MME (w/o subtitle)83.3%estimated ± 1.7 pp, medium confidence
Video-MME (with subtitle)85.3%estimated ± 3.7 pp, medium confidence
VideoMMMU87.2%estimated ± 10.3 pp, low confidence
We-Math87.4%estimated ± 6.7 pp, low confidence
AA LiveCodeBench87.6%estimated ± 4.7 pp, low confidence
HealthBench Hard16.6%estimated ± 8.6 pp, low confidence
LiveCodeBench Pro80.8%estimated ± 6.0 pp, low confidence
MedXpertQA (MM)72.6%estimated ± 6.5 pp, low confidence
MMLU-Pro (Arcee)85.8%estimated ± 4.0 pp, medium confidence
React Native Evals74.0%estimated ± 3.2 pp, high confidence
ResearchClawBench18.2%estimated ± 1.3 pp, high confidence
SWE-Rebench0.1%estimated ± 8.9 pp, low confidence
PostTrainBench21.7%estimated ± 3.7 pp, medium confidence
Finance Agent v255.1%estimated ± 2.4 pp, low confidence
BioMysteryBench (human-solvable)81.7%estimated ± 0.6 pp, low confidence
FrontierCode 1.1 Extended54.7%estimated ± 3.0 pp, low confidence
HealthBench (raw)29.4%estimated ± 6.0 pp, low confidence
HealthBench (length-adjusted)46.5%estimated ± 3.4 pp, low confidence
HealthBench Professional56.7%estimated ± 4.2 pp, low confidence
HealthBench Professional (raw)41.3%estimated ± 4.5 pp, medium confidence
HLE-Verified20.6%estimated ± 7.1 pp, low confidence
ScarfBench24.4%estimated ± 9.1 pp, low confidence
STEMBench2.4%estimated ± 0.2 pp, low confidence
SimpleQA26.1%estimated ± 4.3 pp, medium confidence
Agents' Last Exam29.6%estimated ± 9.7 pp, low confidence
BabyVision w/ Python87.8%estimated ± 4.0 pp, low confidence
OpenHarmony Bench v1.051.0%estimated ± 2.4 pp, medium confidence
MRCRv282.9%estimated ± 11.3 pp, low confidence
MMLongBench-Doc61.1%estimated ± 2.6 pp, low confidence
BioBasicsBench33.9%estimated ± 2.8 pp, low confidence
LifeVal66.4%estimated ± 1.6 pp, medium confidence
MMMU-Pro w/ Python78.5%estimated ± 2.0 pp, high confidence
ExploitBench Internal Port0.1%estimated ± 13.8 pp, low confidence
PinchBench90.0%estimated ± 3.6 pp, medium confidence