Other · model
Hy4 benchmark scores
As of 2026-11-04, Hy4 (Other) has measured scores on 12 benchmarks and estimated scores on 72 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 90.8% | estimated ± 3.0 pp, medium confidence |
| AA-HLE | 44.8% | estimated ± 7.4 pp, medium confidence |
| AA-SciCode | 58.7% | estimated ± 3.5 pp, medium confidence |
| CritPt | 19.6% | estimated ± 6.0 pp, medium confidence |
| GDPval-AA | 64.7% | estimated ± 4.2 pp, high confidence |
| GPQA Diamond | 92.9% | estimated ± 2.3 pp, high confidence |
| HLE | 59.1% | measured |
| SciCode | 61.8% | estimated ± 7.6 pp, low confidence |
| BrowseComp | 92.0% | estimated ± 5.5 pp, low confidence |
| HLE w/ tools | 59.1% | measured |
| τ²-bench results | 98.3% | estimated ± 3.1 pp, high confidence |
| AA Coding Index | 78.1% | estimated ± 2.4 pp, low confidence |
| AA-MMMU-Pro | 77.6% | estimated ± 1.6 pp, high confidence |
| APEX-Agents | 75.2% | measured |
| SWE-bench Verified | 94.0% | estimated ± 4.6 pp, high confidence |
| AutomationBench | 56.2% | measured |
| CyberGym | 88.3% | estimated ± 4.6 pp, medium confidence |
| SWE-bench Pro | 77.5% | estimated ± 7.4 pp, low confidence |
| WideResearch | 83.1% | estimated ± 4.8 pp, low confidence |
| AA Agentic Index | 46.2% | estimated ± 9.2 pp, medium confidence |
| Artificial Analysis Intelligence Index | 39.4% | estimated ± 9.3 pp, low confidence |
| GPQA | 92.2% | estimated ± 3.3 pp, high confidence |
| AA-Omniscience Accuracy | 49.2% | estimated ± 9.2 pp, medium confidence |
| MMMU | 85.3% | estimated ± 1.7 pp, high confidence |
| AA Harvey LAB v1.0 | 90.7% | estimated ± 1.7 pp, medium confidence |
| AA Tau3 Banking | 50.9% | estimated ± 3.8 pp, low confidence |
| ARC-AGI-1 | 95.1% | estimated ± 11.9 pp, low confidence |
| CursorBench 3.2 | 30.6% | estimated ± 3.6 pp, low confidence |
| FrontierCode 1.1 Main | 54.3% | estimated ± 3.7 pp, low confidence |
| FrontierSWE v2 | 48.5% | estimated ± 9.4 pp, medium confidence |
| MazeBench | 2.2% | estimated ± 5.6 pp, medium confidence |
| ParseBench | 62.2% | estimated ± 6.2 pp, medium confidence |
| Vals LiveCodeBench | 77.4% | estimated ± 6.1 pp, low confidence |
| Vals MMLU-Pro | 89.2% | estimated ± 2.5 pp, medium confidence |
| Vals SWE-bench | 85.5% | estimated ± 10.6 pp, low confidence |
| Terminal-Bench 2.1 (Vals AI) | 78.4% | estimated ± 5.4 pp, low confidence |
| AA AutomationBench | 55.8% | estimated ± 11.6 pp, low confidence |
| CWE-Bench-AA | 58.8% | estimated ± 5.2 pp, medium confidence |
| AA Cyber Index | 46.3% | estimated ± 9.4 pp, medium confidence |
| DeepsecBench-AA | 23.5% | estimated ± 7.4 pp, medium confidence |
| GDP.pdf | 22.2% | estimated ± 5.4 pp, medium confidence |
| AA Terminal-Bench 4.0 | 41.1% | estimated ± 6.0 pp, medium confidence |
| Bug Hunt Bench | 22.2% | estimated ± 3.2 pp, medium confidence |
| CursorBench 4.0 | 48.4% | estimated ± 2.7 pp, medium confidence |
| DeepSWE | 74.9% | measured |
| FrontierCode 2.0 | 31.0% | measured |
| HLE w/o tools | 47.3% | estimated ± 5.4 pp, medium confidence |
| OSWorld 2.1 | 78.9% | measured |
| PostTrainBench v1.1 | 43.2% | estimated ± 4.5 pp, high confidence |
| ProgramBench | 30.0% | measured |
| Terminal-Bench 4.0 | 44.4% | measured |
| Toolathlon-Verified | 73.5% | estimated ± 5.7 pp, medium confidence |
| ARC-AGI-3 | 13.0% | estimated ± 12.2 pp, low confidence |
| Hermes Bench | 72.8% | estimated ± 4.7 pp, medium confidence |
| Hermes Index | 39.3% | estimated ± 8.1 pp, low confidence |
| CharXiv | 81.0% | estimated ± 5.2 pp, medium confidence |
| CharXiv w/o tools | 82.4% | estimated ± 2.5 pp, medium confidence |
| ExploitGym | 23.5% | estimated ± 10.5 pp, low confidence |
| AI2D_TEST | 92.6% | estimated ± 2.0 pp, medium confidence |
| CountBench | 95.1% | estimated ± 3.0 pp, high confidence |
| DynaMath | 85.0% | estimated ± 5.3 pp, low confidence |
| ERQA | 63.6% | estimated ± 6.4 pp, medium confidence |
| MathVision | 87.6% | estimated ± 2.9 pp, high confidence |
| MCP Atlas | 79.0% | estimated ± 12.1 pp, low confidence |
| MLVU (M-Avg) | 86.2% | estimated ± 2.2 pp, high confidence |
| MMLU-Redux | 97.1% | estimated ± 2.1 pp, medium confidence |
| MMMU-Pro | 78.6% | measured |
| NL2Repo | 62.1% | estimated ± 6.2 pp, medium confidence |
| RealWorldQA | 83.8% | estimated ± 2.9 pp, high confidence |
| ScreenSpot Pro | 75.5% | estimated ± 10.5 pp, low confidence |
| SimpleVQA | 67.0% | estimated ± 7.5 pp, medium confidence |
| SuperGPQA | 97.4% | estimated ± 6.9 pp, low confidence |
| Video-MME (w/o subtitle) | 84.8% | estimated ± 1.4 pp, medium confidence |
| Video-MME (with subtitle) | 87.1% | estimated ± 3.0 pp, high confidence |
| VideoMMMU | 86.5% | estimated ± 5.1 pp, medium confidence |
| We-Math | 82.5% | estimated ± 7.4 pp, low confidence |
| MedXpertQA (MM) | 71.5% | estimated ± 2.4 pp, high confidence |
| ScarfBench | 69.9% | estimated ± 3.8 pp, medium confidence |
| Agents' Last Exam | 41.0% | measured |
| MMVU | 77.5% | estimated ± 2.7 pp, medium confidence |
| SWE-Marathon | 43.8% | measured |
| MMMU-Pro w/ Python | 81.0% | estimated ± 2.6 pp, high confidence |
| OCRBench V2 | 72.7% | estimated ± 4.4 pp, medium confidence |
| SWE-Atlas Codebase QnA | 64.0% | estimated ± 5.4 pp, low confidence |