OpenAI · model
GPT-6.2 Astra benchmark scores
As of 2026-11-18, GPT-6.2 Astra (OpenAI) has measured scores on 34 benchmarks and estimated scores on 95 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 96.2% | estimated ± 2.1 pp, low confidence |
| AA-HLE | 61.4% | estimated ± 3.9 pp, medium confidence |
| AA-SciCode | 61.7% | estimated ± 3.5 pp, medium confidence |
| AIME26 | 98.2% | estimated ± 1.4 pp, medium confidence |
| Artificial Analysis Intelligence Index | 62.1% | measured |
| CritPt | 35.3% | estimated ± 4.0 pp, medium confidence |
| GDPval-AA | 61.1% | estimated ± 6.9 pp, low confidence |
| GPQA Diamond | 97.0% | measured |
| HLE | 74.5% | estimated ± 2.2 pp, medium confidence |
| AA-Omniscience Accuracy | 68.6% | measured |
| Terminal-Bench 2.1 | 89.1% | estimated ± 5.4 pp, low confidence |
| BrowseComp | 92.7% | estimated ± 2.9 pp, low confidence |
| HLE w/ tools | 64.6% | estimated ± 5.1 pp, low confidence |
| AA Coding Index | 79.0% | estimated ± 2.3 pp, low confidence |
| AA-MMMU-Pro | 89.5% | measured |
| BioMysteryBench (human-difficult) | 51.4% | estimated ± 7.8 pp, low confidence |
| AutomationBench | 52.3% | estimated ± 6.1 pp, low confidence |
| JobBench | 73.6% | estimated ± 11.3 pp, low confidence |
| SWE-bench Pro | 85.9% | estimated ± 8.4 pp, medium confidence |
| AA Agentic Index | 52.2% | estimated ± 12.7 pp, low confidence |
| MMLU-Pro | 85.5% | estimated ± 5.1 pp, low confidence |
| FailureSensorIQ | 69.4% | estimated ± 2.1 pp, medium confidence |
| GPQA | 97.0% | measured |
| MMLU | 88.1% | estimated ± 4.3 pp, medium confidence |
| Vals GPQA Diamond | 98.8% | estimated ± 1.2 pp, low confidence |
| Vals MMLU-Pro | 90.3% | estimated ± 1.6 pp, medium confidence |
| MMMLU | 91.8% | estimated ± 1.1 pp, low confidence |
| MMMU | 91.9% | estimated ± 2.7 pp, medium confidence |
| Chess Puzzles | 78.0% | measured |
| AA MMLU-Pro | 91.6% | estimated ± 0.8 pp, medium confidence |
| AA-AnalystAgent | 62.5% | measured |
| AA EnterpriseOps-Gym | 29.9% | measured |
| AA Tau3 Banking | 100.0% | estimated ± 2.4 pp, low confidence |
| ARC-AGI-1 | 98.5% | measured |
| ARC-AGI-2 | 95.8% | measured |
| BabaIsBench | 100.0% | estimated ± 1.2 pp, medium confidence |
| CursorBench 3.2 | 82.5% | estimated ± 4.1 pp, low confidence |
| EBR-bench | 92.4% | measured |
| FrontierCode 1.1 Main | 51.8% | estimated ± 3.9 pp, medium confidence |
| FrontierMath v2 (Tier 4) | 100.0% | measured |
| FrontierMath v2 (Tiers 1-3) | 92.6% | measured |
| FrontierSWE v2 | 60.6% | estimated ± 8.3 pp, medium confidence |
| Furniture Assembly | 89.2% | measured |
| MaxxBench | 76.8% | estimated ± 4.3 pp, low confidence |
| MazeBench | 9.5% | estimated ± 6.2 pp, low confidence |
| MirrorCode | 67.8% | measured |
| Mystery Game Puzzles | 95.0% | measured |
| NYT Connections | 95.0% | estimated ± 6.7 pp, low confidence |
| OfficeQA Pro | 67.6% | estimated ± 6.8 pp, low confidence |
| ParseBench-1 | 78.0% | estimated ± 6.2 pp, low confidence |
| Terminal-Bench 4.1 | 65.0% | estimated ± 5.5 pp, low confidence |
| Terminal-Bench 2.1 (Vals AI) | 84.2% | estimated ± 6.9 pp, low confidence |
| CWE-Bench-AA | 65.0% | measured |
| AA Cyber Index | 37.3% | measured |
| DeepsecBench-AA | 46.8% | measured |
| Harvey LAB-AA v1.1 | 15.0% | measured |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 64.8% | estimated ± 4.3 pp, medium confidence |
| HealthBench Professional (raw, Artificial Analysis) | 71.4% | estimated ± 1.1 pp, medium confidence |
| Bug Hunt Bench | 47.5% | estimated ± 8.9 pp, medium confidence |
| Chartography (tools) | 88.9% | estimated ± 4.1 pp, low confidence |
| CursorBench 4.0 | 57.6% | estimated ± 3.3 pp, high confidence |
| Epoch Automation Reports | 65.7% | estimated ± 3.0 pp, low confidence |
| FrontierCode 2.0 | 44.2% | measured |
| FrontierPhysics | 20.6% | estimated ± 2.3 pp, high confidence |
| HLE w/o tools | 51.5% | estimated ± 4.2 pp, medium confidence |
| OSWorld 2.0 | 72.5% | estimated ± 14.7 pp, low confidence |
| OSWorld 2.1 Offline | 98.7% | estimated ± 7.9 pp, low confidence |
| PostTrainBench v1.1 | 46.0% | estimated ± 4.8 pp, high confidence |
| SpireClimb | 41.3% | measured |
| StealthMark | 68.6% | estimated ± 6.1 pp, medium confidence |
| Terminal-Bench 4.0 | 67.7% | estimated ± 6.1 pp, medium confidence |
| Terminal-Bench-Science 0.1 | 77.1% | measured |
| Toolathlon-Verified | 79.4% | estimated ± 2.8 pp, low confidence |
| VulcanBench Frontier v4 | 91.7% | estimated ± 6.0 pp, low confidence |
| AA AutomationBench | 77.2% | measured |
| GDP.pdf | 29.8% | measured |
| AA Terminal-Bench 4.0 | 64.1% | measured |
| ARC-AGI-3 | 70.3% | measured |
| ARC-AGI-3 Provider Adapter | 99.8% | measured |
| CWE-bench | 75.5% | estimated ± 6.2 pp, low confidence |
| MKA | 53.2% | estimated ± 5.3 pp, medium confidence |
| SkillsBench official leaderboard (BenchFlow) | 60.8% | estimated ± 6.2 pp, low confidence |
| FACTS Multimodal | 52.4% | estimated ± 2.7 pp, medium confidence |
| Hermes Bench | 100.0% | estimated ± 1.8 pp, low confidence |
| Hermes Index | 92.4% | estimated ± 2.7 pp, low confidence |
| ParseBench-2 | 87.0% | estimated ± 6.2 pp, low confidence |
| CharXiv | 89.5% | estimated ± 6.3 pp, low confidence |
| CharXiv w/o tools | 88.8% | estimated ± 2.5 pp, medium confidence |
| ExploitBench | 100.0% | measured |
| ExploitGym | 41.5% | estimated ± 9.8 pp, low confidence |
| AI2D_TEST | 94.2% | estimated ± 2.0 pp, low confidence |
| C-Eval | 93.9% | estimated ± 1.2 pp, medium confidence |
| ERQA | 68.8% | estimated ± 3.1 pp, medium confidence |
| HMMT Feb 2026 | 100.0% | estimated ± 3.1 pp, medium confidence |
| HMMT Nov 2025 | 96.1% | estimated ± 1.5 pp, low confidence |
| MathVision | 96.9% | estimated ± 3.1 pp, medium confidence |
| MMAnswerBench | 97.4% | estimated ± 2.4 pp, low confidence |
| MMLU-Redux | 95.8% | estimated ± 1.7 pp, medium confidence |
| MMMU-Pro | 87.8% | estimated ± 1.5 pp, medium confidence |
| RealWorldQA | 91.7% | estimated ± 2.9 pp, low confidence |
| ScreenSpot Pro | 82.9% | estimated ± 10.8 pp, low confidence |
| SuperGPQA | 84.7% | estimated ± 8.5 pp, low confidence |
| Video-MME (w/o subtitle) | 93.9% | estimated ± 1.7 pp, low confidence |
| Video-MME (with subtitle) | 93.8% | estimated ± 3.7 pp, low confidence |
| HealthBench Hard | 34.6% | estimated ± 1.2 pp, medium confidence |
| MedXpertQA (MM) | 77.9% | estimated ± 6.5 pp, low confidence |
| MedXpertQA (Text) | 83.5% | estimated ± 4.5 pp, low confidence |
| MMLU-Pro (Arcee) | 89.1% | estimated ± 3.4 pp, low confidence |
| PostTrainBench | 42.0% | estimated ± 4.6 pp, high confidence |
| DAYJOB | 16.6% | estimated ± 1.1 pp, medium confidence |
| DAYJOB Healthcare | 11.2% | estimated ± 1.1 pp, low confidence |
| HealthBench (raw) | 54.7% | estimated ± 1.6 pp, high confidence |
| HealthBench (length-adjusted) | 58.4% | estimated ± 0.9 pp, low confidence |
| HealthBench Professional | 66.3% | measured |
| HealthBench Professional (raw) | 66.0% | measured |
| HLE-Verified | 54.6% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 86.1% | estimated ± 1.7 pp, medium confidence |
| ScarfBench | 69.9% | estimated ± 5.0 pp, low confidence |
| BlenderWorld (text-only) | 86.6% | estimated ± 2.8 pp, low confidence |
| Cookie-Bench | 90.8% | estimated ± 1.3 pp, low confidence |
| STEMBench | 2.7% | estimated ± 0.2 pp, low confidence |
| SimpleQA | 55.0% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 59.1% | estimated ± 11.4 pp, low confidence |
| BabyVision w/ Python | 96.4% | estimated ± 4.0 pp, low confidence |
| SEC-Bench Pro | 86.8% | measured |
| BioBasicsBench | 98.2% | estimated ± 2.2 pp, low confidence |
| LifeVal | 86.1% | measured |
| MMMU-Pro w/ Python | 91.3% | estimated ± 2.0 pp, medium confidence |
| ExploitBench Internal Port | 53.5% | measured |