OpenAI · model
GPT-5.3 Codex benchmark scores
As of 2026-10-07, GPT-5.3 Codex (OpenAI) has measured scores on 19 benchmarks and estimated scores on 80 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 91.5% | measured |
| AA-HLE | 42.5% | measured |
| AA-SciCode | 48.7% | estimated ± 3.8 pp, high confidence |
| Artificial Analysis Intelligence Index | 32.5% | measured |
| CritPt | 16.9% | measured |
| GDPval-AA | 36.8% | estimated ± 6.2 pp, medium confidence |
| GPQA Diamond | 92.3% | estimated ± 2.2 pp, high confidence |
| HLE | 49.6% | estimated ± 9.5 pp, medium confidence |
| IFBench | 74.0% | estimated ± 6.9 pp, medium confidence |
| AA-LCR | 83.3% | measured |
| LiveCodeBench v6 | 87.5% | estimated ± 2.1 pp, high confidence |
| AA-Omniscience Accuracy | 52.9% | measured |
| SciCode | 51.3% | estimated ± 2.9 pp, medium confidence |
| BrowseComp | 74.3% | estimated ± 3.1 pp, high confidence |
| IFEval | 93.7% | estimated ± 1.3 pp, high confidence |
| LongBench v2 | 61.5% | estimated ± 5.9 pp, low confidence |
| τ²-bench results | 86.0% | measured |
| AA Coding Index | 68.6% | estimated ± 5.1 pp, medium confidence |
| AA-MMMU-Pro | 78.5% | measured |
| APEX-Agents-AA | 31.6% | estimated ± 3.6 pp, medium confidence |
| BioMysteryBench (human-difficult) | 46.8% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 75.1% | estimated ± 8.5 pp, low confidence |
| SWE-bench Verified | 85.0% | measured |
| BFCL v4 | 61.6% | estimated ± 9.8 pp, low confidence |
| CyberGym | 56.4% | estimated ± 10.0 pp, low confidence |
| JobBench | 33.7% | measured |
| SWE-bench Pro | 56.8% | measured |
| τ³-bench results | 68.9% | estimated ± 8.9 pp, medium confidence |
| WideResearch | 77.7% | estimated ± 1.2 pp, medium confidence |
| AA Agentic Index | 21.1% | estimated ± 6.4 pp, medium confidence |
| MMLU-Pro | 85.7% | estimated ± 3.5 pp, medium confidence |
| GPQA | 90.8% | estimated ± 3.2 pp, high confidence |
| AA-IFBench | 75.4% | measured |
| Gert Labs | 57.5% | measured |
| AA EnterpriseOps-Gym | 31.3% | estimated ± 3.6 pp, medium confidence |
| ApprenticeBench | 0.0% | estimated ± 3.1 pp, low confidence |
| ARC-AGI-1 | 89.1% | estimated ± 5.4 pp, medium confidence |
| ARC-AGI-2 | 58.9% | estimated ± 14.2 pp, low confidence |
| CursorBench 3.1 | 52.5% | estimated ± 5.6 pp, low confidence |
| CursorBench 3.2 | 53.5% | estimated ± 3.6 pp, high confidence |
| FrontierCode 1.1 Main | 40.7% | estimated ± 2.7 pp, medium confidence |
| FrontierSWE v2 | 9.4% | estimated ± 9.0 pp, medium confidence |
| OfficeQA Pro | 50.7% | estimated ± 7.2 pp, medium confidence |
| OSWorld-Verified | 64.7% | measured |
| Vals GPQA Diamond | 91.4% | estimated ± 7.4 pp, medium confidence |
| Vals LiveCodeBench | 87.3% | measured |
| Vals MMLU-Pro | 89.2% | estimated ± 2.2 pp, high confidence |
| Vals SWE-bench | 78.0% | measured |
| Terminal-Bench 2.1 (Vals AI) | 68.0% | estimated ± 6.3 pp, medium confidence |
| VulcanBench v3 | 85.2% | estimated ± 5.6 pp, low confidence |
| AA AutomationBench | 49.0% | estimated ± 5.0 pp, medium confidence |
| GDP.pdf | 17.8% | estimated ± 5.2 pp, low confidence |
| AA ITBench | 52.2% | estimated ± 3.6 pp, medium confidence |
| CursorBench 4.0 | 33.0% | estimated ± 3.6 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| HLE w/o tools | 41.4% | estimated ± 5.7 pp, medium confidence |
| OSWorld 2.0 | 0.0% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench v1.1 | 23.0% | estimated ± 2.7 pp, high confidence |
| Toolathlon-Verified | 73.2% | estimated ± 1.3 pp, low confidence |
| Vibe Code Bench | 61.8% | measured |
| CharXiv | 85.4% | estimated ± 7.5 pp, low confidence |
| CharXiv w/o tools | 81.8% | estimated ± 2.4 pp, high confidence |
| C-Eval | 94.6% | estimated ± 1.0 pp, low confidence |
| Claw-Eval | 65.5% | estimated ± 5.6 pp, low confidence |
| MathVision | 89.6% | estimated ± 4.9 pp, high confidence |
| MCP Atlas | 58.4% | estimated ± 8.2 pp, medium confidence |
| MCP-Tasks | 72.1% | estimated ± 8.2 pp, low confidence |
| MMLU-Redux | 97.5% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro | 78.5% | estimated ± 4.4 pp, medium confidence |
| NL2Repo | 44.6% | estimated ± 4.7 pp, high confidence |
| ScreenSpot Pro | 77.5% | estimated ± 10.7 pp, low confidence |
| SuperGPQA | 86.8% | estimated ± 7.5 pp, low confidence |
| Terminal-Bench 2.0 | 77.3% | measured |
| Toolathlon | 46.6% | estimated ± 7.2 pp, medium confidence |
| ERQA | 67.8% | estimated ± 3.9 pp, high confidence |
| HealthBench Hard | 29.5% | estimated ± 7.7 pp, low confidence |
| LiveCodeBench Pro | 81.2% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (MM) | 76.3% | estimated ± 6.5 pp, low confidence |
| MMLU-Pro (Arcee) | 100.0% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 78.0% | estimated ± 2.4 pp, medium confidence |
| ResearchClawBench | 17.0% | estimated ± 2.2 pp, medium confidence |
| SWE-Rebench | 58.2% | measured |
| ARC-AGI-3 | 1.7% | estimated ± 3.4 pp, high confidence |
| BioMysteryBench (human-solvable) | 88.4% | estimated ± 0.6 pp, medium confidence |
| FrontierCode 1.1 Extended | 27.1% | estimated ± 2.4 pp, low confidence |
| HealthBench (raw) | 51.4% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 53.9% | estimated ± 3.7 pp, low confidence |
| HealthBench Professional | 57.7% | estimated ± 4.8 pp, medium confidence |
| HealthBench Professional (raw) | 61.6% | estimated ± 5.0 pp, medium confidence |
| HLE-Verified | 38.7% | estimated ± 2.0 pp, medium confidence |
| LABBench2 | 79.4% | estimated ± 1.8 pp, low confidence |
| MMMU | 85.9% | estimated ± 0.9 pp, medium confidence |
| LiveCodeBench | 89.5% | estimated ± 10.2 pp, low confidence |
| Agents' Last Exam | 48.9% | estimated ± 8.7 pp, low confidence |
| OpenHarmony Bench | 53.7% | estimated ± 3.2 pp, high confidence |
| MMLU | 95.3% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro w/ Python | 80.8% | estimated ± 1.6 pp, high confidence |
| RealWorldQA | 86.8% | estimated ± 1.2 pp, medium confidence |
| OmniDocBench 1.5 | 92.0% | estimated ± 6.4 pp, low confidence |