OpenAI · model
GPT-5.4 benchmark scores
As of 2026-10-07, GPT-5.4 (OpenAI) has measured scores on 51 benchmarks and estimated scores on 73 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 92.0% | measured |
| AA-HLE | 43.7% | measured |
| AA-SciCode | 54.3% | estimated ± 3.2 pp, high confidence |
| AIME26 | 100.0% | estimated ± 0.6 pp, low confidence |
| Artificial Analysis Intelligence Index | 39.0% | measured |
| CritPt | 23.4% | measured |
| GDPval-AA | 37.4% | measured |
| GPQA Diamond | 92.8% | measured |
| HLE | 52.1% | measured |
| IFBench | 72.4% | estimated ± 6.9 pp, medium confidence |
| AA-LCR | 82.0% | measured |
| LiveCodeBench v6 | 88.6% | estimated ± 2.1 pp, high confidence |
| AA-Omniscience Accuracy | 50.8% | measured |
| SciCode | 54.0% | estimated ± 4.7 pp, medium confidence |
| BrowseComp | 82.7% | measured |
| HLE w/ tools | 56.6% | estimated ± 4.6 pp, high confidence |
| IFEval | 93.6% | estimated ± 1.3 pp, high confidence |
| LongBench v2 | 61.5% | estimated ± 5.9 pp, low confidence |
| VITA-Bench | 44.3% | estimated ± 9.9 pp, low confidence |
| τ²-bench results | 98.9% | measured |
| AA Coding Index | 71.1% | measured |
| AA-MMMU-Pro | 78.4% | measured |
| APEX-Agents | 26.3% | estimated ± 0.6 pp, medium confidence |
| APEX-Agents-AA | 33.3% | measured |
| BioMysteryBench (human-difficult) | 45.9% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 73.6% | measured |
| SWE-bench Verified | 80.7% | estimated ± 2.5 pp, high confidence |
| AutomationBench | 28.4% | estimated ± 5.8 pp, medium confidence |
| BFCL v4 | 76.2% | estimated ± 2.6 pp, medium confidence |
| CyberGym | 79.0% | measured |
| JobBench | 38.9% | measured |
| SWE-bench Pro | 57.7% | measured |
| τ³-bench results | 69.6% | estimated ± 1.8 pp, medium confidence |
| WideResearch | 78.8% | estimated ± 1.2 pp, medium confidence |
| AA Agentic Index | 33.2% | estimated ± 1.7 pp, high confidence |
| MMLU-Pro | 85.6% | estimated ± 2.5 pp, medium confidence |
| FrontierMath v2 (Tier 4) | 27.1% | measured |
| FrontierMath v2 (Tiers 1-3) | 47.6% | measured |
| GPQA | 92.8% | measured |
| AA-IFBench | 73.9% | measured |
| Gert Labs | 64.9% | measured |
| AA-AnalystAgent | 28.1% | estimated ± 12.0 pp, low confidence |
| AA EnterpriseOps-Gym | 53.5% | estimated ± 3.6 pp, low confidence |
| AA Harvey LAB | 0.1% | estimated ± 2.0 pp, low confidence |
| ApprenticeBench | 11.0% | measured |
| ARC-AGI-1 | 93.7% | measured |
| ARC-AGI-2 | 74.0% | measured |
| CursorBench 3.1 | 50.2% | estimated ± 4.8 pp, medium confidence |
| CursorBench 3.2 | 54.9% | estimated ± 3.6 pp, high confidence |
| FrontierCode 1.1 Main | 40.7% | estimated ± 2.8 pp, low confidence |
| FrontierSWE v2 | 11.0% | estimated ± 9.0 pp, medium confidence |
| OfficeQA Pro | 53.2% | measured |
| OSWorld-Verified | 75.0% | measured |
| Vals GPQA Diamond | 92.2% | estimated ± 1.1 pp, high confidence |
| Vals LiveCodeBench | 86.2% | estimated ± 3.8 pp, low confidence |
| Vals MMLU-Pro | 88.1% | estimated ± 1.2 pp, high confidence |
| Vals SWE-bench | 82.3% | estimated ± 1.4 pp, low confidence |
| Terminal-Bench 2.1 (Vals AI) | 66.0% | estimated ± 6.3 pp, medium confidence |
| AA AutomationBench | 50.4% | estimated ± 5.0 pp, medium confidence |
| GDP.pdf | 22.4% | estimated ± 4.4 pp, low confidence |
| AA ITBench | 39.0% | estimated ± 3.6 pp, low confidence |
| AA Tau3 Banking | 22.0% | estimated ± 5.8 pp, medium confidence |
| CursorBench 4.0 | 33.9% | estimated ± 1.6 pp, medium confidence |
| CWE-bench v1 | 46.5% | estimated ± 6.9 pp, medium confidence |
| DeepSWE | 50.0% | estimated ± 4.4 pp, medium confidence |
| HLE w/o tools | 39.8% | measured |
| OSWorld 2.0 | 9.2% | estimated ± 3.6 pp, high confidence |
| PostTrainBench v1.1 | 19.0% | measured |
| ProgramBench | 0.0% | estimated ± 4.8 pp, low confidence |
| Toolathlon-Verified | 73.2% | estimated ± 1.3 pp, low confidence |
| Vibe Code Bench | 67.4% | measured |
| CharXiv | 82.8% | measured |
| CharXiv w/o tools | 85.7% | estimated ± 1.8 pp, medium confidence |
| ExploitGym | 6.0% | measured |
| C-Eval | 93.3% | estimated ± 0.8 pp, low confidence |
| Claw-Eval | 60.3% | measured |
| DeepPlanning | 45.5% | estimated ± 13.1 pp, low confidence |
| HMMT Nov 2025 | 93.2% | estimated ± 2.6 pp, low confidence |
| MathVision | 89.2% | estimated ± 1.4 pp, low confidence |
| MCP Atlas | 70.6% | measured |
| MCP-Tasks | 76.9% | estimated ± 0.6 pp, low confidence |
| MMAnswerBench | 88.0% | estimated ± 1.3 pp, low confidence |
| MMLU-Redux | 97.4% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro | 81.2% | measured |
| NL2Repo | 52.5% | estimated ± 5.1 pp, medium confidence |
| QwenClawBench | 59.8% | estimated ± 2.6 pp, high confidence |
| ScreenSpot Pro | 85.4% | measured |
| SuperGPQA | 86.0% | estimated ± 7.5 pp, low confidence |
| Terminal-Bench 2.0 | 75.1% | measured |
| Toolathlon | 54.6% | measured |
| VideoMMMU | 87.8% | estimated ± 1.0 pp, high confidence |
| ERQA | 65.4% | measured |
| HealthBench Hard | 40.1% | measured |
| LiveCodeBench Pro | 87.5% | measured |
| MedXpertQA (MM) | 77.1% | measured |
| MedXpertQA (Text) | 59.6% | measured |
| MMLU-Pro (Arcee) | 87.8% | estimated ± 3.4 pp, low confidence |
| React Native Evals | 85.3% | measured |
| ResearchClawBench | 15.3% | measured |
| SWE-Rebench | 62.2% | estimated ± 6.5 pp, low confidence |
| ARC-AGI-3 | 0.2% | measured |
| FrontierMath (legacy) | 48.1% | estimated ± 0.7 pp, low confidence |
| BioMysteryBench (human-solvable) | 89.4% | estimated ± 0.6 pp, medium confidence |
| DRACO | 76.9% | estimated ± 3.0 pp, medium confidence |
| FrontierCode 1.1 Extended | 55.2% | estimated ± 3.0 pp, low confidence |
| HealthBench (raw) | 52.4% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 53.8% | estimated ± 0.9 pp, medium confidence |
| HealthBench Professional | 48.1% | measured |
| HealthBench Professional (raw) | 62.6% | estimated ± 5.0 pp, medium confidence |
| HLE-Verified | 35.3% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 78.0% | estimated ± 1.3 pp, low confidence |
| MMMU | 85.9% | estimated ± 0.9 pp, medium confidence |
| LiveCodeBench | 100.0% | estimated ± 8.9 pp, low confidence |
| Agents' Last Exam | 24.0% | estimated ± 2.1 pp, high confidence |
| OpenHarmony Bench | 56.9% | estimated ± 2.8 pp, high confidence |
| SimpleVQA | 61.1% | measured |
| ZeroBench | 41.0% | measured |
| MMMLU | 90.7% | estimated ± 1.1 pp, low confidence |
| MMLU | 95.0% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro w/ Python | 82.1% | measured |
| MathVision w/ Python | 76.8% | estimated ± 0.9 pp, low confidence |
| RealWorldQA | 85.7% | estimated ± 1.1 pp, medium confidence |
| Video-MME (with subtitle) | 88.0% | estimated ± 0.9 pp, medium confidence |
| OmniDocBench 1.5 | 91.7% | estimated ± 5.9 pp, low confidence |