OpenAI · model
GPT-5.4 mini benchmark scores
As of 2026-10-07, GPT-5.4 mini (OpenAI) has measured scores on 34 benchmarks and estimated scores on 89 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 87.5% | measured |
| AA-HLE | 28.1% | measured |
| AA-SciCode | 52.1% | measured |
| AIME26 | 95.5% | estimated ± 0.6 pp, low confidence |
| Artificial Analysis Intelligence Index | 24.1% | measured |
| CritPt | 10.0% | measured |
| GDPval-AA | 25.8% | measured |
| GPQA Diamond | 88.0% | estimated ± 0.1 pp, high confidence |
| HLE | 41.5% | measured |
| IFBench | 71.8% | estimated ± 6.9 pp, medium confidence |
| AA-LCR | 77.0% | measured |
| LiveCodeBench v6 | 87.7% | estimated ± 3.1 pp, high confidence |
| AA-Omniscience Accuracy | 37.5% | measured |
| SciCode | 39.7% | estimated ± 2.9 pp, low confidence |
| Terminal-Bench 2.1 | 71.0% | estimated ± 6.9 pp, medium confidence |
| BrowseComp | 83.7% | estimated ± 3.1 pp, high confidence |
| HLE w/ tools | 44.8% | estimated ± 6.1 pp, medium confidence |
| IFEval | 93.6% | estimated ± 1.3 pp, high confidence |
| LongBench v2 | 61.5% | estimated ± 5.9 pp, medium confidence |
| VITA-Bench | 38.5% | estimated ± 9.9 pp, low confidence |
| τ²-bench results | 93.4% | measured |
| AA Coding Index | 56.1% | measured |
| AA-MMMU-Pro | 73.3% | measured |
| APEX-Agents | 16.0% | estimated ± 0.6 pp, low confidence |
| APEX-Agents-AA | 28.2% | measured |
| BioMysteryBench (human-difficult) | 39.6% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 84.0% | estimated ± 4.5 pp, high confidence |
| SWE-bench Verified | 80.2% | estimated ± 0.8 pp, medium confidence |
| AutomationBench | 28.4% | estimated ± 9.7 pp, low confidence |
| BFCL v4 | 74.3% | estimated ± 2.6 pp, high confidence |
| CyberGym | 60.6% | estimated ± 5.3 pp, low confidence |
| JobBench | 30.7% | estimated ± 6.8 pp, medium confidence |
| SWE-bench Pro | 57.1% | estimated ± 2.1 pp, high confidence |
| τ³-bench results | 69.5% | estimated ± 1.8 pp, medium confidence |
| WideResearch | 76.0% | estimated ± 5.6 pp, low confidence |
| AA Agentic Index | 19.7% | measured |
| MMLU-Pro | 85.1% | estimated ± 2.5 pp, medium confidence |
| FrontierMath v2 (Tier 4) | 2.1% | measured |
| FrontierMath v2 (Tiers 1-3) | 28.3% | measured |
| GPQA | 88.0% | measured |
| AA-IFBench | 73.3% | measured |
| Gert Labs | 50.7% | estimated ± 7.2 pp, medium confidence |
| AA-AnalystAgent | 18.1% | estimated ± 12.0 pp, low confidence |
| AA EnterpriseOps-Gym | 43.7% | estimated ± 3.6 pp, medium confidence |
| AA Harvey LAB | 80.8% | estimated ± 1.2 pp, low confidence |
| ApprenticeBench | 3.5% | estimated ± 5.0 pp, low confidence |
| ARC-AGI-1 | 63.7% | measured |
| ARC-AGI-2 | 18.9% | measured |
| CursorBench 3.1 | 49.2% | estimated ± 3.4 pp, medium confidence |
| CursorBench 3.2 | 42.5% | estimated ± 1.6 pp, low confidence |
| Terminal-Bench 3.0 | 3.3% | estimated ± 6.9 pp, low confidence |
| FrontierCode 1.1 Main | 27.0% | measured |
| FrontierSWE v2 | 7.3% | estimated ± 9.5 pp, low confidence |
| OfficeQA Pro | 50.3% | estimated ± 7.2 pp, low confidence |
| OSWorld-Verified | 72.1% | measured |
| Vals GPQA Diamond | 83.1% | measured |
| Vals LiveCodeBench | 81.5% | measured |
| Vals MMLU-Pro | 84.6% | measured |
| Vals SWE-bench | 73.0% | measured |
| Terminal-Bench 2.1 (Vals AI) | 54.7% | measured |
| VulcanBench v3 | 81.6% | estimated ± 5.6 pp, low confidence |
| AA AutomationBench | 4.3% | estimated ± 7.0 pp, medium confidence |
| GDP.pdf | 9.4% | estimated ± 5.4 pp, medium confidence |
| AA ITBench | 29.0% | estimated ± 3.6 pp, low confidence |
| AA Tau3 Banking | 22.0% | estimated ± 5.8 pp, medium confidence |
| AA Terminal-Bench 2.1 | 68.7% | estimated ± 4.4 pp, high confidence |
| AA Terminal-Bench 4.0 | 7.7% | estimated ± 14.0 pp, low confidence |
| CursorBench 4.0 | 19.8% | estimated ± 1.6 pp, medium confidence |
| CWE-bench v1 | 35.7% | estimated ± 6.9 pp, low confidence |
| DeepSWE | 22.2% | estimated ± 3.4 pp, low confidence |
| HLE w/o tools | 28.2% | measured |
| OSWorld 2.0 | 5.9% | estimated ± 3.6 pp, high confidence |
| PostTrainBench v1.1 | 21.8% | estimated ± 2.7 pp, high confidence |
| Toolathlon-Verified | 47.7% | estimated ± 2.3 pp, low confidence |
| Vibe Code Bench | 48.0% | measured |
| CharXiv | 79.2% | estimated ± 5.4 pp, medium confidence |
| CharXiv w/o tools | 80.7% | estimated ± 1.8 pp, medium confidence |
| ExploitGym | 0.0% | estimated ± 12.2 pp, low confidence |
| C-Eval | 93.2% | estimated ± 0.8 pp, low confidence |
| Claw-Eval | 63.9% | estimated ± 5.7 pp, medium confidence |
| DeepPlanning | 38.8% | estimated ± 13.1 pp, low confidence |
| HMMT Nov 2025 | 96.8% | estimated ± 2.6 pp, low confidence |
| MathVision | 83.7% | estimated ± 3.6 pp, high confidence |
| MCP Atlas | 57.7% | measured |
| MCP-Tasks | 76.2% | estimated ± 0.6 pp, low confidence |
| MMAnswerBench | 82.4% | estimated ± 1.3 pp, low confidence |
| MMLU-Redux | 96.4% | estimated ± 1.1 pp, medium confidence |
| MMMU-Pro | 76.6% | measured |
| NL2Repo | 42.0% | estimated ± 4.7 pp, high confidence |
| QwenClawBench | 55.2% | estimated ± 2.6 pp, high confidence |
| ScreenSpot Pro | 69.7% | estimated ± 10.0 pp, low confidence |
| SuperGPQA | 79.7% | estimated ± 7.5 pp, medium confidence |
| Terminal-Bench 2.0 | 60.0% | measured |
| Toolathlon | 42.9% | measured |
| VideoMMMU | 84.4% | estimated ± 1.0 pp, high confidence |
| ERQA | 56.4% | estimated ± 3.9 pp, high confidence |
| HealthBench Hard | 0.0% | estimated ± 5.1 pp, low confidence |
| MedXpertQA (MM) | 67.0% | estimated ± 2.4 pp, high confidence |
| MedXpertQA (Text) | 44.3% | estimated ± 8.9 pp, low confidence |
| MMLU-Pro (Arcee) | 100.0% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 70.1% | estimated ± 2.4 pp, low confidence |
| ResearchClawBench | 17.0% | estimated ± 1.8 pp, medium confidence |
| SWE-Rebench | 62.0% | estimated ± 6.5 pp, low confidence |
| ARC-AGI-3 | 0.0% | estimated ± 2.1 pp, medium confidence |
| FrontierMath (legacy) | 29.0% | estimated ± 0.7 pp, low confidence |
| BioMysteryBench (human-solvable) | 83.3% | estimated ± 0.6 pp, low confidence |
| DRACO | 72.1% | estimated ± 3.0 pp, medium confidence |
| FrontierCode 1.1 Extended | 19.6% | estimated ± 2.4 pp, low confidence |
| HealthBench (raw) | 37.2% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 43.7% | estimated ± 2.5 pp, low confidence |
| HealthBench Professional | 51.7% | estimated ± 1.3 pp, low confidence |
| HealthBench Professional (raw) | 48.4% | estimated ± 5.0 pp, low confidence |
| HLE-Verified | 35.3% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 22.8% | estimated ± 1.3 pp, low confidence |
| MMMU | 81.6% | estimated ± 0.9 pp, high confidence |
| LiveCodeBench | 100.0% | estimated ± 8.9 pp, low confidence |
| OpenHarmony Bench | 58.6% | estimated ± 2.4 pp, high confidence |
| SimpleVQA | 66.2% | estimated ± 8.2 pp, medium confidence |
| MMLU | 92.8% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro w/ Python | 78.0% | measured |
| RealWorldQA | 84.0% | estimated ± 1.2 pp, low confidence |
| Video-MME (with subtitle) | 86.9% | estimated ± 1.8 pp, medium confidence |
| OmniDocBench 1.5 | 91.5% | estimated ± 5.9 pp, low confidence |