Google · model
Gemini 4 Neon benchmark scores
As of 2026-11-19, Gemini 4 Neon (Google) has measured scores on 6 benchmarks and estimated scores on 51 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 91.9% | estimated ± 11.8 pp, low confidence |
| AA-HLE | 58.9% | estimated ± 9.2 pp, medium confidence |
| AA-SciCode | 63.1% | estimated ± 3.5 pp, medium confidence |
| Artificial Analysis Intelligence Index | 61.4% | measured |
| CritPt | 32.2% | estimated ± 5.3 pp, medium confidence |
| GDPval-AA | 60.9% | estimated ± 6.4 pp, low confidence |
| GPQA Diamond | 93.1% | estimated ± 2.5 pp, high confidence |
| HLE | 65.5% | estimated ± 12.0 pp, low confidence |
| AA-Omniscience Accuracy | 59.0% | estimated ± 7.9 pp, medium confidence |
| Terminal-Bench 2.1 | 89.1% | estimated ± 5.4 pp, low confidence |
| HLE w/ tools | 66.7% | estimated ± 6.8 pp, low confidence |
| AA Coding Index | 79.1% | estimated ± 2.3 pp, low confidence |
| BioMysteryBench (human-difficult) | 49.8% | estimated ± 8.3 pp, low confidence |
| AutomationBench | 53.4% | estimated ± 6.1 pp, low confidence |
| JobBench | 72.7% | estimated ± 11.3 pp, low confidence |
| SWE-bench Pro | 87.5% | estimated ± 8.4 pp, medium confidence |
| MMLU-Pro | 86.3% | estimated ± 5.5 pp, low confidence |
| GPQA | 93.1% | estimated ± 6.9 pp, medium confidence |
| Vals GPQA Diamond | 93.1% | estimated ± 5.7 pp, low confidence |
| Vals MMLU-Pro | 90.1% | estimated ± 4.2 pp, medium confidence |
| AA-AnalystAgent | 58.1% | estimated ± 6.3 pp, low confidence |
| ARC-AGI-1 | 100.0% | estimated ± 7.9 pp, low confidence |
| ARC-AGI-2 | 90.1% | estimated ± 14.6 pp, low confidence |
| CursorBench 3.2 | 83.4% | estimated ± 4.1 pp, low confidence |
| FrontierCode 1.1 Main | 52.3% | estimated ± 3.9 pp, medium confidence |
| FrontierSWE v2 | 62.5% | estimated ± 8.3 pp, medium confidence |
| MazeBench | 17.0% | estimated ± 6.8 pp, low confidence |
| Terminal-Bench 4.1 | 62.6% | estimated ± 6.0 pp, medium confidence |
| Terminal-Bench 2.1 (Vals AI) | 84.0% | estimated ± 6.9 pp, low confidence |
| CWE-Bench-AA | 63.3% | estimated ± 3.8 pp, medium confidence |
| DeepsecBench-AA | 47.6% | estimated ± 7.5 pp, low confidence |
| Harvey LAB-AA v1.1 | 9.2% | estimated ± 3.5 pp, medium confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 64.6% | estimated ± 4.3 pp, medium confidence |
| HealthBench Professional (raw, Artificial Analysis) | 71.2% | estimated ± 1.1 pp, medium confidence |
| Bug Hunt Bench | 51.4% | estimated ± 8.8 pp, medium confidence |
| CursorBench 4.0 | 58.7% | estimated ± 3.3 pp, high confidence |
| FrontierCode 2.0 | 46.2% | measured |
| HLE w/o tools | 65.0% | estimated ± 5.0 pp, high confidence |
| OSWorld 2.1 Offline | 83.8% | estimated ± 7.4 pp, medium confidence |
| PostTrainBench v1.1 | 47.1% | estimated ± 4.8 pp, high confidence |
| Terminal-Bench 4.0 | 67.2% | estimated ± 6.1 pp, medium confidence |
| Toolathlon-Verified | 78.0% | estimated ± 1.8 pp, low confidence |
| VulcanBench Frontier v4 | 92.4% | estimated ± 6.0 pp, low confidence |
| AA AutomationBench | 79.3% | measured |
| GDP.pdf | 28.0% | measured |
| AA Terminal-Bench 4.0 | 63.6% | measured |
| CWE-bench | 75.6% | estimated ± 5.0 pp, medium confidence |
| MKA | 56.5% | measured |
| Hermes Bench | 100.0% | estimated ± 1.8 pp, low confidence |
| Hermes Index | 90.8% | estimated ± 2.7 pp, low confidence |
| ExploitGym | 50.0% | estimated ± 7.2 pp, low confidence |
| PostTrainBench | 42.8% | estimated ± 4.6 pp, high confidence |
| HealthBench (length-adjusted) | 60.7% | estimated ± 3.7 pp, medium confidence |
| HealthBench Professional | 65.9% | estimated ± 4.2 pp, medium confidence |
| ScarfBench | 69.9% | estimated ± 5.0 pp, low confidence |
| Agents' Last Exam | 52.3% | estimated ± 10.7 pp, low confidence |
| BioBasicsBench | 58.4% | estimated ± 10.2 pp, low confidence |