Google · model
Gemini 4 Argon benchmark scores
As of 2026-10-07, Gemini 4 Argon (Google) has measured scores on 20 benchmarks and estimated scores on 67 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 96.6% | estimated ± 1.2 pp, low confidence |
| AA-HLE | 57.1% | measured |
| AA-SciCode | 61.8% | measured |
| Artificial Analysis Intelligence Index | 52.6% | measured |
| CritPt | 27.1% | measured |
| GDPval-AA | 56.3% | measured |
| GPQA Diamond | 97.1% | estimated ± 2.2 pp, medium confidence |
| HLE | 55.5% | estimated ± 9.5 pp, medium confidence |
| AA-LCR | 79.7% | measured |
| LiveCodeBench v6 | 100.0% | estimated ± 3.3 pp, medium confidence |
| AA-Omniscience Accuracy | 49.9% | measured |
| SciCode | 61.1% | estimated ± 3.2 pp, medium confidence |
| Terminal-Bench 2.1 | 89.7% | estimated ± 3.9 pp, medium confidence |
| BrowseComp | 91.5% | estimated ± 2.2 pp, high confidence |
| HLE w/ tools | 60.5% | estimated ± 6.1 pp, medium confidence |
| LongBench v2 | 61.5% | estimated ± 5.9 pp, low confidence |
| AA Coding Index | 78.6% | estimated ± 2.5 pp, medium confidence |
| APEX-Agents | 45.1% | estimated ± 10.0 pp, low confidence |
| APEX-Agents-AA | 38.9% | estimated ± 2.9 pp, low confidence |
| BioMysteryBench (human-difficult) | 45.6% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 90.4% | estimated ± 8.4 pp, medium confidence |
| SWE-bench Verified | 89.0% | estimated ± 4.4 pp, medium confidence |
| AutomationBench | 51.3% | measured |
| BFCL v4 | 77.9% | estimated ± 2.6 pp, medium confidence |
| CyberGym | 88.8% | estimated ± 3.7 pp, high confidence |
| JobBench | 65.7% | estimated ± 4.9 pp, low confidence |
| SWE-bench Pro | 81.5% | estimated ± 5.1 pp, medium confidence |
| WideResearch | 100.0% | estimated ± 9.1 pp, low confidence |
| AA Agentic Index | 51.1% | estimated ± 2.8 pp, high confidence |
| MMLU-Pro | 85.7% | estimated ± 3.5 pp, medium confidence |
| GPQA | 92.8% | estimated ± 3.2 pp, high confidence |
| Gert Labs | 76.7% | estimated ± 3.7 pp, low confidence |
| AA-AnalystAgent | 54.4% | estimated ± 5.5 pp, medium confidence |
| AA EnterpriseOps-Gym | 46.9% | estimated ± 4.4 pp, high confidence |
| AA Harvey LAB | 93.3% | estimated ± 1.3 pp, medium confidence |
| ARC-AGI-1 | 99.7% | estimated ± 0.7 pp, low confidence |
| ARC-AGI-2 | 83.0% | estimated ± 14.2 pp, low confidence |
| CursorBench 3.1 | 73.1% | estimated ± 3.4 pp, low confidence |
| CursorBench 3.2 | 72.4% | estimated ± 4.1 pp, low confidence |
| FrontierCode 1.1 Main | 53.8% | estimated ± 2.8 pp, medium confidence |
| FrontierSWE v2 | 55.0% | measured |
| OSWorld-Verified | 100.0% | estimated ± 3.0 pp, medium confidence |
| Vals GPQA Diamond | 96.4% | estimated ± 2.2 pp, low confidence |
| Vals MMLU-Pro | 92.8% | estimated ± 1.4 pp, low confidence |
| Vals SWE-bench | 100.0% | estimated ± 4.5 pp, medium confidence |
| Terminal-Bench 2.1 (Vals AI) | 86.3% | estimated ± 4.6 pp, medium confidence |
| AA AutomationBench | 77.5% | measured |
| GDP.pdf | 21.8% | measured |
| AA Tau3 Banking | 47.5% | estimated ± 5.8 pp, medium confidence |
| AA Terminal-Bench 2.1 | 86.8% | estimated ± 2.6 pp, high confidence |
| AA Terminal-Bench 4.0 | 57.1% | measured |
| CursorBench 4.0 | 49.9% | estimated ± 2.9 pp, high confidence |
| CWE-bench v1 | 68.0% | measured |
| DeepSWE | 77.9% | measured |
| HLE w/o tools | 61.3% | estimated ± 5.7 pp, medium confidence |
| OSWorld 2.0 | 69.2% | measured |
| PostTrainBench v1.1 | 45.3% | measured |
| ProgramBench | 90.9% | estimated ± 4.8 pp, medium confidence |
| Terminal-Bench 4.0 | 57.4% | measured |
| Terminal-Bench-Science 0.1 | 58.4% | estimated ± 4.1 pp, medium confidence |
| Toolathlon-Verified | 79.8% | estimated ± 1.0 pp, medium confidence |
| Vibe Code Bench | 91.9% | measured |
| ExploitGym | 17.3% | estimated ± 5.0 pp, low confidence |
| C-Eval | 96.8% | estimated ± 1.0 pp, low confidence |
| MCP Atlas | 89.2% | estimated ± 2.7 pp, low confidence |
| MMLU-Redux | 97.4% | estimated ± 1.1 pp, low confidence |
| NL2Repo | 68.6% | estimated ± 6.5 pp, low confidence |
| QwenClawBench | 100.0% | estimated ± 6.0 pp, low confidence |
| SuperGPQA | 85.7% | estimated ± 7.5 pp, low confidence |
| Toolathlon | 55.9% | estimated ± 6.3 pp, low confidence |
| HealthBench Hard | 35.0% | estimated ± 7.7 pp, low confidence |
| MMLU-Pro (Arcee) | 100.0% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 86.5% | estimated ± 3.0 pp, low confidence |
| ResearchClawBench | 19.9% | estimated ± 2.2 pp, low confidence |
| SWE-Rebench | 62.3% | estimated ± 6.5 pp, low confidence |
| ARC-AGI-3 | 1.7% | estimated ± 3.4 pp, high confidence |
| BioMysteryBench (human-solvable) | 89.5% | estimated ± 0.6 pp, medium confidence |
| DRACO | 84.6% | estimated ± 3.0 pp, medium confidence |
| HealthBench (raw) | 63.4% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 59.5% | estimated ± 3.7 pp, medium confidence |
| HealthBench Professional | 64.9% | estimated ± 4.8 pp, medium confidence |
| HealthBench Professional (raw) | 72.9% | estimated ± 5.0 pp, medium confidence |
| HLE-Verified | 54.8% | estimated ± 1.2 pp, low confidence |
| LABBench2 | 88.8% | measured |
| Agents' Last Exam | 39.5% | measured |
| OpenHarmony Bench | 58.8% | estimated ± 2.4 pp, medium confidence |
| MMLU | 94.9% | estimated ± 1.1 pp, low confidence |