Other · model
Gemma 4 26B A4B benchmark scores
As of 2026-11-22, Gemma 4 26B A4B (Other) has measured scores on 23 benchmarks and estimated scores on 128 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 79.2% | measured |
| AA-HLE | 19.3% | measured |
| AA-SciCode | 40.0% | measured |
| Artificial Analysis Intelligence Index | 0.0% | estimated ± 5.6 pp, low confidence |
| CritPt | 0.0% | measured |
| GDPval-AA | 3.2% | measured |
| GPQA Diamond | 82.4% | estimated ± 1.2 pp, high confidence |
| HLE | 17.2% | measured |
| IFBench | 70.4% | estimated ± 6.6 pp, medium confidence |
| AA-LCR | 65.7% | measured |
| LiveCodeBench v6 | 77.1% | measured |
| AA-Omniscience Accuracy | 18.4% | measured |
| SciCode | 40.1% | estimated ± 3.1 pp, high confidence |
| BrowseComp | 64.0% | estimated ± 8.8 pp, low confidence |
| HLE w/ tools | 35.3% | estimated ± 5.3 pp, low confidence |
| IFEval | 92.9% | estimated ± 2.8 pp, high confidence |
| LongBench v2 | 56.4% | estimated ± 5.5 pp, medium confidence |
| τ²-bench results | 43.6% | measured |
| AA Coding Index | 39.3% | measured |
| AA-MMMU-Pro | 69.2% | measured |
| APEX-Agents | 2.0% | estimated ± 9.4 pp, low confidence |
| APEX-Agents-AA | 8.8% | estimated ± 9.3 pp, low confidence |
| BioMysteryBench (human-difficult) | 0.0% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 43.8% | estimated ± 9.6 pp, medium confidence |
| SWE-bench Verified | 65.9% | estimated ± 6.3 pp, medium confidence |
| BFCL v4 | 60.5% | estimated ± 2.7 pp, high confidence |
| CyberGym | 8.8% | estimated ± 6.0 pp, low confidence |
| JobBench | 0.0% | estimated ± 8.7 pp, low confidence |
| SWE-bench Pro | 32.8% | estimated ± 6.0 pp, medium confidence |
| WideResearch | 61.2% | estimated ± 3.3 pp, medium confidence |
| AA Agentic Index | 2.2% | estimated ± 3.1 pp, high confidence |
| MMLU-Pro | 82.6% | measured |
| AA-IFBench | 72.4% | measured |
| BBH | 97.4% | estimated ± 2.8 pp, medium confidence |
| GPQA | 82.3% | measured |
| MMLU | 81.6% | estimated ± 0.9 pp, high confidence |
| Vals GPQA Diamond | 57.5% | estimated ± 2.9 pp, medium confidence |
| Vals LiveCodeBench | 76.7% | estimated ± 5.7 pp, low confidence |
| Vals MMLU-Pro | 61.1% | estimated ± 1.8 pp, medium confidence |
| MMMLU | 83.3% | estimated ± 0.7 pp, medium confidence |
| MMMU | 80.4% | estimated ± 0.8 pp, high confidence |
| Chess Puzzles | 23.1% | estimated ± 10.8 pp, low confidence |
| AA MMLU-Pro | 83.8% | estimated ± 0.8 pp, medium confidence |
| Gert Labs | 19.6% | measured |
| AA Tau3 Banking | 5.8% | estimated ± 4.1 pp, medium confidence |
| ApprenticeBench | 3.1% | estimated ± 10.4 pp, low confidence |
| ARC-AGI-1 | 13.7% | estimated ± 3.9 pp, medium confidence |
| ARC-AGI-2 | 4.2% | estimated ± 10.9 pp, low confidence |
| CursorBench 3.2 | 37.7% | estimated ± 5.3 pp, low confidence |
| EBR-bench | 14.8% | estimated ± 11.3 pp, low confidence |
| FrontierCode 1.1 Main | 13.6% | estimated ± 4.6 pp, medium confidence |
| FrontierSWE v2 | 5.5% | estimated ± 14.2 pp, low confidence |
| Furniture Assembly | 33.4% | estimated ± 12.4 pp, low confidence |
| MaxxBench | 37.6% | estimated ± 5.7 pp, low confidence |
| MazeBench | 0.0% | estimated ± 2.7 pp, medium confidence |
| Mystery Game Puzzles | 11.4% | estimated ± 13.3 pp, low confidence |
| NYT Connections | 73.2% | estimated ± 6.3 pp, medium confidence |
| OfficeQA Pro | 50.3% | estimated ± 6.8 pp, low confidence |
| OSWorld-Verified | 57.1% | estimated ± 8.4 pp, low confidence |
| ParseBench-1 | 58.5% | measured |
| Vals SWE-bench | 62.3% | estimated ± 5.5 pp, medium confidence |
| CWE-Bench-AA | 30.7% | estimated ± 8.1 pp, low confidence |
| AA Cyber Index | 2.8% | estimated ± 12.4 pp, low confidence |
| DeepsecBench-AA | 5.7% | estimated ± 8.2 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 4.3% | estimated ± 2.1 pp, low confidence |
| AA ITBench | 35.6% | estimated ± 4.6 pp, low confidence |
| Bug Hunt Bench | 14.6% | estimated ± 10.5 pp, low confidence |
| Chartography (tools) | 76.4% | estimated ± 4.1 pp, low confidence |
| CursorBench 4.0 | 26.6% | estimated ± 4.9 pp, medium confidence |
| CWE-bench v1 | 32.6% | estimated ± 5.7 pp, low confidence |
| DeepSWE | 31.9% | estimated ± 7.5 pp, low confidence |
| DRACO | 25.2% | estimated ± 2.0 pp, low confidence |
| FrontierCode 2.0 | 26.1% | estimated ± 9.3 pp, low confidence |
| FrontierPhysics | 1.0% | estimated ± 4.7 pp, medium confidence |
| HLE w/o tools | 8.7% | measured |
| OSWorld 2.1 | 1.0% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench v1.1 | 11.0% | estimated ± 6.2 pp, low confidence |
| StealthMark | 21.2% | measured |
| SWE-bench Multimodal | 9.2% | estimated ± 5.6 pp, low confidence |
| Toolathlon-Verified | 54.1% | estimated ± 3.0 pp, high confidence |
| VulcanBench Frontier v4 | 69.7% | estimated ± 6.8 pp, low confidence |
| AA AutomationBench | 37.5% | estimated ± 9.9 pp, low confidence |
| GDP.pdf | 9.4% | estimated ± 4.2 pp, medium confidence |
| ARC-AGI-3 | 0.0% | estimated ± 7.2 pp, low confidence |
| ARC-AGI-3 Provider Adapter | 2.7% | estimated ± 7.3 pp, low confidence |
| CWE-bench | 25.2% | estimated ± 6.2 pp, low confidence |
| MKA | 32.7% | estimated ± 4.6 pp, medium confidence |
| Vibe Code Bench | 14.2% | estimated ± 10.0 pp, low confidence |
| FACTS Multimodal | 39.5% | measured |
| Hermes Bench | 14.7% | estimated ± 2.9 pp, medium confidence |
| Hermes Index | 14.1% | estimated ± 5.7 pp, low confidence |
| ParseBench-2 | 62.6% | estimated ± 3.5 pp, high confidence |
| CharXiv | 83.2% | estimated ± 5.6 pp, low confidence |
| CharXiv w/o tools | 78.3% | estimated ± 1.2 pp, high confidence |
| ExploitBench | 50.2% | estimated ± 7.3 pp, low confidence |
| ExploitGym | 0.0% | estimated ± 8.3 pp, low confidence |
| AI2D_TEST | 88.1% | estimated ± 2.0 pp, medium confidence |
| C-Eval | 87.8% | estimated ± 0.9 pp, low confidence |
| CountBench | 90.9% | estimated ± 2.9 pp, low confidence |
| DynaMath | 59.5% | estimated ± 3.0 pp, low confidence |
| ERQA | 42.7% | estimated ± 3.1 pp, medium confidence |
| MathVision | 82.6% | estimated ± 3.0 pp, high confidence |
| MCP Atlas | 56.5% | estimated ± 10.7 pp, low confidence |
| MLVU (M-Avg) | 84.0% | estimated ± 2.2 pp, high confidence |
| MMLU-Redux | 91.9% | estimated ± 0.6 pp, high confidence |
| MMMU-Pro | 73.8% | measured |
| NL2Repo | 25.4% | estimated ± 3.4 pp, medium confidence |
| QwenClawBench | 55.9% | estimated ± 6.0 pp, low confidence |
| RealWorldQA | 82.8% | estimated ± 2.8 pp, high confidence |
| ScreenSpot Pro | 72.8% | estimated ± 9.6 pp, low confidence |
| SimpleVQA | 61.3% | estimated ± 7.5 pp, medium confidence |
| SuperGPQA | 58.4% | estimated ± 7.2 pp, medium confidence |
| Toolathlon | 27.0% | estimated ± 6.3 pp, low confidence |
| V* | 39.3% | estimated ± 7.5 pp, low confidence |
| Video-MME (w/o subtitle) | 82.2% | estimated ± 1.3 pp, high confidence |
| Video-MME (with subtitle) | 84.9% | estimated ± 3.0 pp, high confidence |
| VideoMMMU | 82.5% | estimated ± 2.4 pp, high confidence |
| We-Math | 80.4% | estimated ± 7.4 pp, low confidence |
| AA LiveCodeBench | 87.3% | estimated ± 4.7 pp, low confidence |
| HealthBench Hard | 0.0% | estimated ± 8.6 pp, low confidence |
| MedXpertQA (MM) | 60.3% | estimated ± 2.4 pp, high confidence |
| MMLU-Pro (Arcee) | 77.1% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 74.0% | estimated ± 3.2 pp, high confidence |
| ResearchClawBench | 11.7% | estimated ± 2.2 pp, low confidence |
| SWE-Rebench | 0.0% | estimated ± 8.9 pp, low confidence |
| PostTrainBench | 9.9% | estimated ± 4.1 pp, low confidence |
| DAYJOB | 0.0% | estimated ± 3.2 pp, medium confidence |
| DAYJOB Healthcare | 0.0% | estimated ± 4.7 pp, medium confidence |
| Finance Agent v2 | 55.1% | estimated ± 2.4 pp, low confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| HealthBench (raw) | 26.9% | estimated ± 6.0 pp, low confidence |
| HealthBench (length-adjusted) | 45.7% | estimated ± 3.4 pp, low confidence |
| HealthBench Professional | 58.1% | estimated ± 4.5 pp, low confidence |
| HealthBench Professional (raw) | 39.0% | estimated ± 4.5 pp, medium confidence |
| ScarfBench | 24.4% | measured |
| STEMBench | 2.3% | estimated ± 0.2 pp, low confidence |
| SimpleQA | 15.9% | estimated ± 4.3 pp, medium confidence |
| Agents' Last Exam | 29.6% | estimated ± 9.7 pp, low confidence |
| BabyVision w/ Python | 88.8% | estimated ± 3.8 pp, low confidence |
| OpenHarmony Bench v1.0 | 51.0% | estimated ± 2.4 pp, medium confidence |
| SEC-Bench Pro | 39.5% | estimated ± 5.6 pp, medium confidence |
| BabyVision | 50.9% | estimated ± 9.1 pp, low confidence |
| MMVU | 74.6% | estimated ± 2.7 pp, medium confidence |
| PerceptionBench | 57.4% | estimated ± 4.9 pp, low confidence |
| MRCRv2 | 44.1% | measured |
| MMLongBench-Doc | 60.7% | estimated ± 2.6 pp, low confidence |
| OCRBench V2 | 70.0% | estimated ± 3.3 pp, high confidence |
| BioBasicsBench | 34.1% | estimated ± 3.8 pp, low confidence |
| LifeVal | 0.6% | estimated ± 5.0 pp, low confidence |
| MMMU-Pro w/ Python | 73.0% | estimated ± 2.0 pp, high confidence |
| ExploitBench Internal Port | 0.0% | estimated ± 6.8 pp, low confidence |