xAI · model
Grok 4.20 benchmark scores
As of 2026-10-07, Grok 4.20 (xAI) has measured scores on 23 benchmarks and estimated scores on 66 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 88.7% | estimated ± 1.2 pp, medium confidence |
| AA-HLE | 40.7% | estimated ± 2.6 pp, medium confidence |
| AA-SciCode | 46.1% | estimated ± 3.8 pp, high confidence |
| Artificial Analysis Intelligence Index | 26.9% | estimated ± 5.1 pp, medium confidence |
| CritPt | 7.8% | estimated ± 4.9 pp, high confidence |
| GDPval-AA | 15.6% | estimated ± 8.8 pp, medium confidence |
| GPQA Diamond | 88.5% | measured |
| HLE | 44.8% | estimated ± 5.5 pp, low confidence |
| LiveCodeBench v6 | 85.6% | estimated ± 2.1 pp, high confidence |
| AA-Omniscience Accuracy | 47.6% | estimated ± 4.6 pp, medium confidence |
| SciCode | 45.3% | estimated ± 2.9 pp, medium confidence |
| Terminal-Bench 2.1 | 46.8% | estimated ± 6.9 pp, low confidence |
| BrowseComp | 58.7% | estimated ± 5.7 pp, medium confidence |
| AA Coding Index | 53.8% | estimated ± 5.1 pp, medium confidence |
| AA-MMMU-Pro | 73.2% | estimated ± 2.2 pp, high confidence |
| APEX-Agents-AA | 10.5% | estimated ± 11.4 pp, low confidence |
| DeepSearchQA | 62.8% | measured |
| SWE-bench Verified | 76.7% | measured |
| CyberGym | 30.1% | estimated ± 10.0 pp, low confidence |
| SWE-bench Pro | 51.8% | measured |
| τ³-bench results | 97.5% | estimated ± 8.9 pp, low confidence |
| WideResearch | 54.0% | estimated ± 5.9 pp, low confidence |
| AA Agentic Index | 10.2% | estimated ± 6.4 pp, medium confidence |
| MMLU-Pro | 85.3% | estimated ± 2.5 pp, medium confidence |
| GPQA | 88.5% | estimated ± 0.1 pp, high confidence |
| Gert Labs | 38.4% | measured |
| ApprenticeBench | 0.0% | estimated ± 3.1 pp, low confidence |
| ARC-AGI-1 | 86.7% | estimated ± 1.9 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 |
| Terminal-Bench 3.0 | 0.9% | estimated ± 6.9 pp, low confidence |
| FrontierCode 1.1 Main | 14.9% | estimated ± 2.7 pp, low confidence |
| FrontierSWE v2 | 0.5% | estimated ± 9.0 pp, low confidence |
| OSWorld-Verified | 53.8% | estimated ± 5.9 pp, medium confidence |
| Vals GPQA Diamond | 88.6% | measured |
| Vals LiveCodeBench | 84.3% | measured |
| Vals MMLU-Pro | 86.3% | measured |
| 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 |
| AA Terminal-Bench 2.1 | 42.1% | estimated ± 4.4 pp, medium confidence |
| AA Terminal-Bench 4.0 | 0.0% | estimated ± 14.0 pp, low confidence |
| CursorBench 4.0 | 29.2% | estimated ± 3.6 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| HLE w/o tools | 31.6% | measured |
| OSWorld 2.0 | 0.0% | estimated ± 4.5 pp, low confidence |
| PostTrainBench v1.1 | 20.4% | estimated ± 2.7 pp, high confidence |
| Vibe Code Bench | 4.1% | measured |
| CharXiv | 60.9% | measured |
| CharXiv w/o tools | 79.2% | estimated ± 1.8 pp, medium confidence |
| Claw-Eval | 54.5% | estimated ± 6.8 pp, low confidence |
| MathVision | 86.6% | estimated ± 1.4 pp, low confidence |
| MCP Atlas | 43.5% | estimated ± 14.8 pp, low confidence |
| MMLU-Redux | 92.8% | estimated ± 3.0 pp, medium confidence |
| MMMU-Pro | 75.2% | measured |
| NL2Repo | 41.6% | estimated ± 4.7 pp, high confidence |
| ScreenSpot Pro | 85.0% | estimated ± 2.5 pp, low confidence |
| SuperGPQA | 73.9% | estimated ± 13.9 pp, low confidence |
| Terminal-Bench 2.0 | 47.1% | measured |
| Toolathlon | 33.7% | estimated ± 7.2 pp, low confidence |
| VideoMMMU | 84.2% | estimated ± 1.0 pp, high confidence |
| ERQA | 54.1% | measured |
| 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 | 68.9% | estimated ± 2.4 pp, low confidence |
| ResearchClawBench | 14.0% | estimated ± 2.2 pp, medium confidence |
| SWE-Rebench | 59.0% | estimated ± 3.6 pp, medium confidence |
| ARC-AGI-3 | 0.1% | measured |
| FrontierCode 1.1 Extended | 18.6% | estimated ± 2.4 pp, low confidence |
| HealthBench (length-adjusted) | 53.8% | estimated ± 0.9 pp, low confidence |
| HealthBench Professional | 56.6% | estimated ± 1.3 pp, medium confidence |
| HealthBench Professional (raw) | 62.0% | estimated ± 5.3 pp, low confidence |
| HLE-Verified | 35.3% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 81.1% | estimated ± 1.8 pp, low confidence |
| MMMU | 82.7% | estimated ± 1.1 pp, medium confidence |
| LiveCodeBench | 84.4% | estimated ± 10.2 pp, low confidence |
| OpenHarmony Bench | 51.8% | estimated ± 3.2 pp, high confidence |
| SimpleVQA | 57.4% | measured |
| ZeroBench | 46.2% | estimated ± 6.4 pp, low confidence |
| MMMLU | 89.3% | estimated ± 1.1 pp, medium confidence |
| MMMU-Pro w/ Python | 76.7% | estimated ± 0.6 pp, high confidence |
| MathVision w/ Python | 76.8% | estimated ± 0.9 pp, low confidence |
| RealWorldQA | 77.3% | estimated ± 1.1 pp, low confidence |
| Video-MME (with subtitle) | 83.5% | estimated ± 0.9 pp, low confidence |
| OmniDocBench 1.5 | 89.5% | estimated ± 5.9 pp, low confidence |