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.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 93.2% | estimated ± 4.4 pp, high confidence |
| AA-HLE | 46.5% | measured |
| AA-SciCode | 58.9% | measured |
| Artificial Analysis Intelligence Index | 28.5% | measured |
| CritPt | 20.9% | measured |
| GDPval-AA | 54.3% | measured |
| GPQA Diamond | 93.5% | measured |
| HLE | 46.5% | measured |
| AA-LCR | 88.3% | measured |
| LiveCodeBench v6 | 93.1% | estimated ± 4.4 pp, high confidence |
| AA-Omniscience Accuracy | 36.4% | measured |
| SciCode | 58.9% | measured |
| Terminal-Bench 2.1 | 85.0% | measured |
| BrowseComp | 88.7% | measured |
| HLE w/ tools | 55.0% | estimated ± 3.1 pp, high confidence |
| LongBench v2 | 73.6% | estimated ± 5.5 pp, low confidence |
| VITA-Bench | 46.9% | estimated ± 6.7 pp, low confidence |
| τ²-bench results | 95.4% | estimated ± 7.0 pp, medium confidence |
| AA Coding Index | 75.5% | estimated ± 2.5 pp, high confidence |
| AA-MMMU-Pro | 76.4% | measured |
| APEX-Agents | 37.8% | measured |
| APEX-Agents-AA | 38.0% | measured |
| BioMysteryBench (human-difficult) | 48.3% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 91.2% | estimated ± 6.8 pp, low confidence |
| SWE-bench Verified | 87.0% | estimated ± 5.1 pp, medium confidence |
| AutomationBench | 44.0% | measured |
| BFCL v4 | 77.0% | estimated ± 2.7 pp, medium confidence |
| CyberGym | 84.7% | measured |
| JobBench | 59.0% | measured |
| SkillsBench | 61.5% | estimated ± 4.1 pp, medium confidence |
| SWE-bench Pro | 73.8% | estimated ± 6.3 pp, medium confidence |
| WideResearch | 83.2% | estimated ± 1.2 pp, medium confidence |
| AA Agentic Index | 50.0% | estimated ± 3.1 pp, high confidence |
| MMLU-Pro | 85.5% | estimated ± 5.0 pp, medium confidence |
| FailureSensorIQ | 67.1% | estimated ± 1.7 pp, high confidence |
| GPQA | 93.5% | estimated ± 0.0 pp, high confidence |
| MMLU | 90.5% | estimated ± 0.9 pp, medium confidence |
| Vals GPQA Diamond | 92.5% | estimated ± 2.1 pp, high confidence |
| Vals LiveCodeBench | 88.5% | estimated ± 0.8 pp, low confidence |
| Vals MMLU-Pro | 88.5% | estimated ± 1.6 pp, high confidence |
| MMMLU | 90.9% | estimated ± 0.7 pp, low confidence |
| MMMU | 83.8% | estimated ± 0.8 pp, high confidence |
| Chess Puzzles | 38.8% | estimated ± 10.8 pp, low confidence |
| AA MMLU-Pro | 86.6% | estimated ± 0.8 pp, high confidence |
| Gert Labs | 69.7% | estimated ± 5.5 pp, low confidence |
| AA-AnalystAgent | 53.3% | estimated ± 3.7 pp, medium confidence |
| AA EnterpriseOps-Gym | 47.2% | measured |
| AA Harvey LAB v1.0 | 93.4% | measured |
| AA Tau3 Banking | 47.1% | estimated ± 2.4 pp, low confidence |
| AA Terminal-Bench 2.1 | 83.0% | estimated ± 2.6 pp, high confidence |
| ApprenticeBench | 26.6% | estimated ± 4.0 pp, medium confidence |
| ARC-AGI-1 | 94.2% | estimated ± 2.5 pp, high confidence |
| ARC-AGI-2 | 76.1% | estimated ± 9.7 pp, medium confidence |
| CursorBench 3.1 | 65.9% | estimated ± 3.4 pp, medium confidence |
| CursorBench 3.2 | 65.3% | estimated ± 3.6 pp, medium confidence |
| EBR-bench | 30.3% | estimated ± 11.3 pp, low confidence |
| Terminal-Bench 3.0 | 18.3% | estimated ± 9.1 pp, low confidence |
| FrontierCode 1.1 Main | 48.6% | estimated ± 3.5 pp, low confidence |
| FrontierSWE v2 | 39.2% | estimated ± 14.2 pp, low confidence |
| Furniture Assembly | 36.2% | estimated ± 12.4 pp, low confidence |
| MaxxBench | 49.3% | estimated ± 5.7 pp, medium confidence |
| MazeBench | 2.7% | estimated ± 4.0 pp, high confidence |
| Mystery Game Puzzles | 25.3% | estimated ± 13.3 pp, low confidence |
| NYT Connections | 76.6% | measured |
| OfficeQA Pro | 60.3% | measured |
| OSWorld-Verified | 80.0% | estimated ± 4.7 pp, high confidence |
| ParseBench-1 | 67.9% | estimated ± 5.4 pp, low confidence |
| Terminal-Bench 4.1 | 25.3% | estimated ± 4.6 pp, high confidence |
| Vals SWE-bench | 77.8% | estimated ± 5.3 pp, low confidence |
| Terminal-Bench 2.1 (Vals AI) | 71.4% | estimated ± 7.3 pp, medium confidence |
| CWE-Bench-AA | 52.5% | measured |
| AA Cyber Index | 47.7% | measured |
| DeepsecBench-AA | 15.8% | 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 ITBench | 55.6% | measured |
| Bug Hunt Bench | 62.4% | estimated ± 3.2 pp, low confidence |
| Chartography (tools) | 81.1% | estimated ± 4.1 pp, low confidence |
| CursorBench 4.0 | 53.9% | estimated ± 4.7 pp, high confidence |
| CWE-bench v1 | 58.6% | estimated ± 6.4 pp, medium confidence |
| DeepSWE | 67.7% | measured |
| DRACO | 83.3% | measured |
| FrontierCode 2.0 | 26.3% | estimated ± 9.3 pp, low confidence |
| FrontierPhysics | 1.7% | estimated ± 4.7 pp, high confidence |
| HLE w/o tools | 39.4% | estimated ± 2.7 pp, high confidence |
| OfficeQA | 73.9% | estimated ± 0.2 pp, low confidence |
| OSWorld 2.0 | 19.4% | estimated ± 5.5 pp, low confidence |
| OSWorld 2.1 | 69.5% | estimated ± 6.0 pp, medium confidence |
| OSWorld 2.1 Offline | 69.6% | estimated ± 1.9 pp, medium confidence |
| PostTrainBench v1.1 | 35.6% | estimated ± 4.5 pp, high confidence |
| ProgramBench | 80.5% | measured |
| SpireClimb | 5.8% | estimated ± 7.2 pp, medium confidence |
| StealthMark | 39.0% | measured |
| SWE-bench Multimodal | 36.7% | estimated ± 8.8 pp, low confidence |
| Terminal-Bench 4.0 | 33.3% | measured |
| Terminal-Bench-Science 0.1 | 7.2% | estimated ± 14.2 pp, low confidence |
| Toolathlon-Verified | 74.1% | measured |
| VulcanBench Frontier v4 | 92.2% | estimated ± 1.2 pp, medium confidence |
| AA AutomationBench | 51.7% | estimated ± 5.5 pp, medium confidence |
| GDP.pdf | 8.6% | estimated ± 3.1 pp, medium confidence |
| AA Terminal-Bench 4.0 | 30.8% | estimated ± 6.3 pp, low confidence |
| ARC-AGI-3 | 0.0% | estimated ± 8.3 pp, low confidence |
| ARC-AGI-3 Provider Adapter | 11.6% | estimated ± 7.3 pp, medium confidence |
| cursorBench41 | 45.1% | estimated ± 3.7 pp, high confidence |
| CWE-bench | 42.0% | estimated ± 4.3 pp, low confidence |
| MKA | 27.6% | measured |
| SkillsBench official leaderboard (BenchFlow) | 53.7% | estimated ± 6.1 pp, medium confidence |
| BinaryAudit | 35.8% | estimated ± 12.1 pp, low confidence |
| OTelBench | 22.5% | estimated ± 6.6 pp, medium confidence |
| FACTS Multimodal | 43.3% | estimated ± 4.0 pp, high confidence |
| Hermes Bench | 69.2% | measured |
| Hermes Index | 32.8% | measured |
| ParseBench-2 | 66.2% | estimated ± 5.0 pp, medium confidence |
| CharXiv | 87.0% | estimated ± 5.1 pp, low confidence |
| CharXiv w/o tools | 80.2% | estimated ± 1.2 pp, high confidence |
| ExploitBench | 26.1% | estimated ± 12.6 pp, low confidence |
| ExploitGym | 9.2% | estimated ± 6.4 pp, low confidence |
| AI2D_TEST | 93.1% | estimated ± 2.0 pp, medium confidence |
| C-Eval | 94.1% | estimated ± 1.0 pp, low confidence |
| Claw-Eval | 75.6% | estimated ± 4.5 pp, medium confidence |
| CountBench | 95.9% | estimated ± 2.9 pp, medium confidence |
| DeepPlanning | 52.6% | estimated ± 13.1 pp, low confidence |
| DynaMath | 85.4% | estimated ± 3.0 pp, medium confidence |
| ERQA | 66.5% | estimated ± 3.1 pp, high confidence |
| MathVision | 85.1% | estimated ± 3.0 pp, high confidence |
| MCP Atlas | 85.6% | measured |
| MLVU (M-Avg) | 85.0% | estimated ± 2.2 pp, high confidence |
| MMLU-Redux | 95.5% | estimated ± 0.8 pp, medium confidence |
| MMMU-Pro | 76.0% | measured |
| NL2Repo | 59.8% | estimated ± 6.0 pp, medium confidence |
| QwenClawBench | 70.7% | estimated ± 2.6 pp, medium confidence |
| RealWorldQA | 83.2% | estimated ± 2.8 pp, high confidence |
| ScreenSpot Pro | 74.1% | estimated ± 9.6 pp, low confidence |
| SimpleVQA | 63.9% | estimated ± 7.5 pp, medium confidence |
| SuperGPQA | 86.0% | estimated ± 7.2 pp, medium confidence |
| Toolathlon | 59.2% | 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 |
| VideoMMMU | 84.1% | estimated ± 2.4 pp, high confidence |
| We-Math | 81.4% | estimated ± 7.4 pp, low confidence |
| HealthBench Hard | 32.7% | 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 Evals | 80.5% | estimated ± 4.7 pp, low confidence |
| ResearchClawBench | 19.9% | estimated ± 1.8 pp, medium confidence |
| DAYJOB | 0.1% | estimated ± 3.2 pp, high confidence |
| DAYJOB Healthcare | 0.0% | 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 Professional | 57.9% | estimated ± 4.2 pp, medium confidence |
| HealthBench Professional (raw) | 64.8% | estimated ± 4.5 pp, high confidence |
| HLE-Verified | 54.3% | estimated ± 6.8 pp, low confidence |
| LABBench2 | 82.6% | estimated ± 2.4 pp, medium confidence |
| ScarfBench | 65.6% | estimated ± 3.8 pp, medium confidence |
| BlenderWorld (text-only) | 72.5% | estimated ± 4.3 pp, medium confidence |
| Cookie-Bench | 83.3% | estimated ± 4.8 pp, medium confidence |
| STEMBench | 2.5% | estimated ± 0.2 pp, medium confidence |
| SimpleQA | 40.6% | estimated ± 4.3 pp, low confidence |
| Agents' Last Exam | 29.5% | measured |
| BabyVision w/ Python | 88.8% | estimated ± 3.8 pp, low confidence |
| OpenHarmony Bench v1.0 | 58.9% | estimated ± 2.4 pp, high confidence |
| SEC-Bench Pro | 45.4% | estimated ± 9.3 pp, medium confidence |
| BabyVision | 56.1% | estimated ± 9.1 pp, low confidence |
| MMVU | 76.0% | estimated ± 2.7 pp, medium confidence |
| PerceptionBench | 58.6% | estimated ± 4.9 pp, low confidence |
| MRCRv2 | 86.0% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 61.7% | estimated ± 2.6 pp, low confidence |
| OCRBench V2 | 70.6% | estimated ± 3.3 pp, high confidence |
| SWE-Marathon | 72.7% | measured |
| BioBasicsBench | 36.1% | estimated ± 2.8 pp, medium confidence |
| LifeVal | 78.8% | estimated ± 1.6 pp, medium confidence |
| MMMU-Pro w/ Python | 79.5% | estimated ± 2.0 pp, high confidence |
| ExploitBench Internal Port | 0.0% | estimated ± 6.8 pp, low confidence |
| PinchBench | 90.6% | estimated ± 3.6 pp, low confidence |
| SWE-Atlas Codebase QnA | 59.9% | estimated ± 4.8 pp, high confidence |