xAI · model
Grok 4.20 benchmark scores
As of 2026-11-22, Grok 4.20 (xAI) has measured scores on 26 benchmarks and estimated scores on 128 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 90.0% | estimated ± 2.7 pp, high confidence |
| AA-HLE | 31.5% | estimated ± 3.9 pp, high confidence |
| AA-SciCode | 46.0% | estimated ± 3.6 pp, high confidence |
| Artificial Analysis Intelligence Index | 21.1% | estimated ± 5.6 pp, medium confidence |
| CritPt | 8.3% | estimated ± 4.6 pp, high confidence |
| GDPval-AA | 16.6% | estimated ± 10.5 pp, low confidence |
| GPQA Diamond | 88.5% | measured |
| HLE | 54.2% | estimated ± 2.2 pp, medium confidence |
| LiveCodeBench v6 | 88.4% | estimated ± 2.2 pp, medium confidence |
| AA-Omniscience Accuracy | 37.3% | estimated ± 6.5 pp, medium confidence |
| SciCode | 45.3% | estimated ± 2.9 pp, medium confidence |
| Terminal-Bench 2.1 | 46.8% | estimated ± 6.7 pp, low confidence |
| BrowseComp | 76.3% | estimated ± 6.8 pp, low confidence |
| HLE w/ tools | 42.9% | estimated ± 5.7 pp, medium confidence |
| AA Coding Index | 54.0% | estimated ± 5.3 pp, medium confidence |
| AA-MMMU-Pro | 74.4% | estimated ± 1.5 pp, high confidence |
| APEX-Agents-AA | 10.3% | estimated ± 9.3 pp, medium confidence |
| DeepSearchQA | 62.8% | measured |
| SWE-bench Verified | 76.7% | measured |
| CyberGym | 44.0% | estimated ± 11.0 pp, low confidence |
| JobBench | 16.0% | estimated ± 14.2 pp, low confidence |
| SWE-bench Pro | 51.8% | measured |
| WideResearch | 64.3% | estimated ± 3.3 pp, high confidence |
| AA Agentic Index | 13.2% | estimated ± 9.4 pp, medium confidence |
| MMLU-Pro | 85.3% | estimated ± 2.5 pp, medium confidence |
| FailureSensorIQ | 67.7% | estimated ± 1.7 pp, high confidence |
| GPQA | 88.5% | estimated ± 0.0 pp, medium confidence |
| MMLU | 93.6% | estimated ± 1.2 pp, medium confidence |
| Vals GPQA Diamond | 88.6% | measured |
| Vals LiveCodeBench | 84.3% | measured |
| Vals MMLU-Pro | 86.3% | measured |
| MMMLU | 89.3% | estimated ± 1.1 pp, medium confidence |
| MMMU | 82.8% | estimated ± 1.7 pp, high confidence |
| Chess Puzzles | 29.0% | measured |
| AA MMLU-Pro | 87.3% | estimated ± 1.1 pp, high confidence |
| Gert Labs | 22.3% | measured |
| AA-AnalystAgent | 44.8% | estimated ± 5.9 pp, low confidence |
| AA EnterpriseOps-Gym | 40.9% | estimated ± 2.3 pp, low confidence |
| AA Terminal-Bench 2.1 | 79.4% | estimated ± 2.5 pp, medium confidence |
| ApprenticeBench | 3.8% | estimated ± 10.1 pp, low confidence |
| ARC-AGI-1 | 86.7% | estimated ± 1.7 pp, high confidence |
| ARC-AGI-2 | 53.3% | measured |
| CursorBench 3.1 | 49.0% | estimated ± 5.6 pp, low confidence |
| CursorBench 3.2 | 44.1% | estimated ± 3.6 pp, medium confidence |
| EBR-bench | 15.5% | estimated ± 12.7 pp, low confidence |
| FrontierCode 1.1 Main | 21.7% | estimated ± 4.2 pp, low confidence |
| FrontierSWE v2 | 1.4% | estimated ± 8.3 pp, low confidence |
| Furniture Assembly | 29.9% | estimated ± 14.1 pp, low confidence |
| MaxxBench | 46.0% | estimated ± 10.3 pp, low confidence |
| MazeBench | 0.0% | measured |
| Mystery Game Puzzles | 20.6% | estimated ± 10.3 pp, low confidence |
| NYT Connections | 85.4% | measured |
| OSWorld-Verified | 63.2% | estimated ± 7.1 pp, low confidence |
| ParseBench-1 | 57.3% | estimated ± 8.4 pp, medium confidence |
| Terminal-Bench 4.1 | 6.2% | estimated ± 14.9 pp, low confidence |
| Terminal-Bench Hard | 44.1% | estimated ± 3.9 pp, medium confidence |
| Vals SWE-bench | 72.2% | measured |
| Terminal-Bench 2.1 (Vals AI) | 44.2% | measured |
| VulcanBench v3 | 83.2% | estimated ± 5.6 pp, low confidence |
| CWE-Bench-AA | 33.4% | estimated ± 8.1 pp, low confidence |
| AA Cyber Index | 5.8% | estimated ± 12.4 pp, low confidence |
| DeepsecBench-AA | 6.5% | estimated ± 8.2 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 57.5% | estimated ± 4.4 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 59.3% | estimated ± 2.6 pp, low confidence |
| AA ITBench | 35.7% | estimated ± 4.6 pp, low confidence |
| Bug Hunt Bench | 27.6% | estimated ± 12.1 pp, low confidence |
| CursorBench 4.0 | 29.4% | estimated ± 3.3 pp, medium confidence |
| CWE-bench v1 | 57.4% | estimated ± 5.1 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| DRACO | 86.9% | estimated ± 0.7 pp, low confidence |
| FrontierCode 2.0 | 19.2% | estimated ± 6.7 pp, low confidence |
| FrontierPhysics | 0.0% | estimated ± 6.1 pp, low confidence |
| HLE w/o tools | 31.6% | measured |
| OSWorld 2.1 | 2.2% | estimated ± 3.6 pp, medium confidence |
| OSWorld 2.1 Offline | 33.3% | estimated ± 13.8 pp, low confidence |
| PostTrainBench v1.1 | 17.3% | estimated ± 3.7 pp, medium confidence |
| SpireClimb | 5.8% | estimated ± 7.2 pp, medium confidence |
| StealthMark | 42.1% | estimated ± 9.7 pp, low confidence |
| Toolathlon-Verified | 73.4% | estimated ± 2.6 pp, high confidence |
| VulcanBench Frontier v4 | 74.7% | estimated ± 8.3 pp, low confidence |
| AA AutomationBench | 39.0% | estimated ± 9.9 pp, low confidence |
| GDP.pdf | 10.5% | estimated ± 4.2 pp, medium confidence |
| AA Terminal-Bench 4.0 | 0.0% | estimated ± 13.7 pp, low confidence |
| ARC-AGI-3 | 0.1% | measured |
| ARC-AGI-3 Provider Adapter | 0.5% | estimated ± 4.4 pp, medium confidence |
| CWE-bench | 27.4% | estimated ± 6.2 pp, low confidence |
| MKA | 32.8% | estimated ± 4.6 pp, medium confidence |
| SkillsBench official leaderboard (BenchFlow) | 54.1% | estimated ± 6.1 pp, medium confidence |
| BinaryAudit | 35.8% | estimated ± 12.1 pp, low confidence |
| OTelBench | 22.5% | estimated ± 6.6 pp, medium confidence |
| Vibe Code Bench | 4.1% | measured |
| FACTS Multimodal | 48.0% | estimated ± 4.0 pp, high confidence |
| Hermes Index | 37.2% | estimated ± 3.1 pp, high confidence |
| ParseBench-2 | 65.1% | estimated ± 5.0 pp, medium confidence |
| CharXiv | 60.9% | measured |
| CharXiv w/o tools | 79.5% | estimated ± 1.2 pp, high confidence |
| ExploitBench | 31.1% | estimated ± 13.4 pp, low confidence |
| ExploitGym | 11.9% | estimated ± 6.3 pp, medium confidence |
| AI2D_TEST | 87.8% | estimated ± 0.9 pp, low confidence |
| CountBench | 93.3% | estimated ± 2.2 pp, medium confidence |
| DynaMath | 72.3% | estimated ± 0.8 pp, low confidence |
| ERQA | 54.1% | measured |
| MathVision | 84.2% | estimated ± 3.0 pp, high confidence |
| MLVU (M-Avg) | 79.4% | estimated ± 1.5 pp, medium confidence |
| MMLU-Redux | 92.4% | estimated ± 2.8 pp, high confidence |
| MMMU-Pro | 75.2% | measured |
| NL2Repo | 41.1% | estimated ± 4.7 pp, high confidence |
| OmniDocBench 1.5 | 87.5% | estimated ± 2.0 pp, medium confidence |
| PostTrainBench v1 | 19.0% | estimated ± 3.1 pp, low confidence |
| RealWorldQA | 74.1% | estimated ± 1.5 pp, medium confidence |
| ScreenSpot Pro | 31.1% | estimated ± 6.8 pp, low confidence |
| SimpleVQA | 57.4% | measured |
| SuperGPQA | 73.9% | estimated ± 13.9 pp, low confidence |
| Terminal-Bench 2.0 | 47.1% | measured |
| Toolathlon | 37.8% | estimated ± 8.0 pp, medium confidence |
| V* | 81.3% | estimated ± 11.2 pp, low confidence |
| Video-MME (w/o subtitle) | 83.0% | estimated ± 1.3 pp, high confidence |
| Video-MME (with subtitle) | 76.5% | estimated ± 2.9 pp, medium confidence |
| VideoMMMU | 83.6% | estimated ± 1.5 pp, low confidence |
| We-Math | 75.2% | estimated ± 2.1 pp, medium confidence |
| AA LiveCodeBench | 87.6% | estimated ± 2.6 pp, high confidence |
| HealthBench Hard | 20.3% | measured |
| LiveCodeBench Pro | 74.2% | measured |
| MedXpertQA (MM) | 65.8% | measured |
| MedXpertQA (Text) | 50.2% | measured |
| MMLU-Pro (Arcee) | 86.4% | estimated ± 3.4 pp, medium confidence |
| React Native Evals | 78.2% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 13.6% | estimated ± 2.9 pp, low confidence |
| SWE-Rebench | 59.0% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench | 14.7% | estimated ± 3.7 pp, medium confidence |
| DAYJOB | 0.5% | estimated ± 6.3 pp, medium confidence |
| DAYJOB Healthcare | 0.2% | estimated ± 6.1 pp, low confidence |
| Finance Agent v2 | 55.1% | estimated ± 2.4 pp, low confidence |
| FrontierCode 1.1 Extended | 53.0% | estimated ± 3.0 pp, low confidence |
| HealthBench (length-adjusted) | 53.8% | estimated ± 0.9 pp, low confidence |
| HealthBench Professional | 56.8% | estimated ± 3.6 pp, medium confidence |
| HLE-Verified | 26.0% | estimated ± 4.6 pp, low confidence |
| LABBench2 | 81.1% | estimated ± 1.8 pp, low confidence |
| ScarfBench | 43.2% | estimated ± 12.3 pp, low confidence |
| BlenderWorld (text-only) | 70.6% | estimated ± 4.0 pp, medium confidence |
| Cookie-Bench | 79.7% | estimated ± 2.5 pp, medium confidence |
| LiveCodeBench | 77.3% | estimated ± 14.6 pp, low 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 | 51.8% | estimated ± 3.2 pp, high confidence |
| SEC-Bench Pro | 41.3% | estimated ± 12.3 pp, low confidence |
| BabyVision | 54.2% | estimated ± 9.1 pp, low confidence |
| MMVU | 75.5% | estimated ± 2.7 pp, medium confidence |
| PerceptionBench | 58.1% | estimated ± 4.9 pp, low confidence |
| OCRBench V2 | 70.4% | estimated ± 3.3 pp, high confidence |
| BioBasicsBench | 36.8% | estimated ± 3.5 pp, low confidence |
| LifeVal | 75.2% | estimated ± 3.8 pp, medium confidence |
| MMMU-Pro w/ Python | 78.8% | estimated ± 2.6 pp, high confidence |
| ExploitBench Internal Port | 0.1% | estimated ± 1.4 pp, low confidence |