DeepSeek · model
DeepSeek V4 Flash 0731 benchmark scores
As of 2026-10-07, DeepSeek V4 Flash 0731 (DeepSeek) has measured scores on 40 benchmarks and estimated scores on 70 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 90.8% | measured |
| AA-HLE | 38.6% | measured |
| AA-SciCode | 50.3% | measured |
| AIME26 | 96.8% | estimated ± 1.3 pp, medium confidence |
| Artificial Analysis Intelligence Index | 34.3% | measured |
| CritPt | 16.6% | measured |
| GDPval-AA | 46.9% | measured |
| GPQA Diamond | 88.1% | measured |
| HLE | 34.8% | measured |
| AA-LCR | 79.7% | measured |
| LiveCodeBench v6 | 85.9% | estimated ± 2.1 pp, high confidence |
| AA-Omniscience Accuracy | 40.4% | measured |
| SciCode | 51.3% | estimated ± 2.9 pp, medium confidence |
| Terminal-Bench 2.1 | 82.7% | measured |
| BrowseComp | 73.2% | measured |
| HLE w/ tools | 45.1% | measured |
| LongBench v2 | 61.5% | estimated ± 5.9 pp, low confidence |
| VITA-Bench | 43.7% | estimated ± 9.9 pp, low confidence |
| τ²-bench results | 94.1% | estimated ± 13.3 pp, low confidence |
| AA Coding Index | 69.1% | measured |
| APEX-Agents | 25.1% | estimated ± 0.6 pp, medium confidence |
| APEX-Agents-AA | 18.8% | estimated ± 4.6 pp, high confidence |
| BioMysteryBench (human-difficult) | 41.1% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 92.1% | estimated ± 4.5 pp, high confidence |
| SWE-bench Verified | 79.0% | measured |
| AutomationBench | 25.1% | measured |
| BFCL v4 | 77.2% | estimated ± 2.6 pp, medium confidence |
| CyberGym | 76.7% | measured |
| JobBench | 50.4% | estimated ± 3.4 pp, low confidence |
| SWE-bench Pro | 52.6% | measured |
| τ³-bench results | 69.6% | estimated ± 1.8 pp, medium confidence |
| WideResearch | 75.0% | estimated ± 3.4 pp, medium confidence |
| AA Agentic Index | 41.7% | measured |
| MMLU-Pro | 86.2% | measured |
| FrontierMath v2 (Tiers 1-3) | 39.0% | estimated ± 8.4 pp, low confidence |
| GPQA | 88.1% | measured |
| Gert Labs | 66.8% | estimated ± 6.0 pp, low confidence |
| AA-AnalystAgent | 17.5% | estimated ± 2.5 pp, medium confidence |
| AA EnterpriseOps-Gym | 49.4% | estimated ± 5.1 pp, low confidence |
| AA Harvey LAB | 92.5% | estimated ± 1.2 pp, medium confidence |
| ApprenticeBench | 6.2% | estimated ± 4.0 pp, medium confidence |
| ARC-AGI-1 | 89.0% | measured |
| ARC-AGI-2 | 61.4% | measured |
| CursorBench 3.1 | 44.7% | estimated ± 3.4 pp, low confidence |
| CursorBench 3.2 | 67.8% | estimated ± 3.3 pp, medium confidence |
| Terminal-Bench 3.0 | 10.6% | estimated ± 6.9 pp, low confidence |
| FrontierCode 1.1 Main | 21.8% | estimated ± 2.7 pp, low confidence |
| FrontierSWE v2 | 2.0% | estimated ± 9.0 pp, low confidence |
| OSWorld-Verified | 78.2% | estimated ± 4.7 pp, high confidence |
| Vals GPQA Diamond | 89.9% | measured |
| Vals LiveCodeBench | 87.3% | measured |
| Vals MMLU-Pro | 86.2% | measured |
| Vals SWE-bench | 88.8% | measured |
| Terminal-Bench 2.1 (Vals AI) | 67.0% | measured |
| VulcanBench v3 | 88.4% | measured |
| AA AutomationBench | 3.5% | estimated ± 5.7 pp, low confidence |
| GDP.pdf | 12.8% | estimated ± 3.4 pp, medium confidence |
| AA ITBench | 60.4% | estimated ± 3.4 pp, low confidence |
| AA Tau3 Banking | 47.2% | estimated ± 5.8 pp, medium confidence |
| AA Terminal-Bench 2.1 | 84.8% | estimated ± 3.5 pp, high confidence |
| AA Terminal-Bench 4.0 | 21.8% | estimated ± 8.9 pp, medium confidence |
| CursorBench 4.0 | 31.5% | estimated ± 1.6 pp, medium confidence |
| CWE-bench v1 | 53.7% | estimated ± 6.9 pp, medium confidence |
| DeepSWE | 54.4% | measured |
| HLE w/o tools | 28.2% | estimated ± 2.9 pp, high confidence |
| OSWorld 2.0 | 19.4% | estimated ± 5.5 pp, low confidence |
| PostTrainBench v1.1 | 29.8% | estimated ± 2.7 pp, high confidence |
| SWE Multilingual | 73.3% | measured |
| Toolathlon-Verified | 70.3% | measured |
| Vibe Code Bench | 35.7% | estimated ± 8.1 pp, medium confidence |
| ExploitGym | 0.0% | estimated ± 6.1 pp, low confidence |
| C-Eval | 93.1% | estimated ± 0.8 pp, low confidence |
| Claw-Eval | 67.6% | estimated ± 5.7 pp, medium confidence |
| DeepPlanning | 44.7% | estimated ± 13.1 pp, low confidence |
| HMMT Feb 2025 | 99.4% | estimated ± 1.8 pp, low confidence |
| HMMT Feb 2026 | 94.8% | measured |
| MCP Atlas | 69.0% | measured |
| MCP-Tasks | 76.9% | estimated ± 0.6 pp, low confidence |
| MMAnswerBench | 87.6% | estimated ± 2.8 pp, low confidence |
| MMLU-ProX | 84.1% | estimated ± 1.7 pp, medium confidence |
| MMLU-Redux | 93.7% | estimated ± 0.7 pp, high confidence |
| NL2Repo | 54.2% | measured |
| NOVA-63 | 56.4% | estimated ± 1.6 pp, low confidence |
| QwenClawBench | 59.1% | estimated ± 2.6 pp, high confidence |
| SuperGPQA | 81.4% | estimated ± 7.5 pp, medium confidence |
| Terminal-Bench 2.0 | 56.9% | measured |
| Toolathlon | 47.8% | measured |
| HealthBench Hard | 0.0% | estimated ± 5.1 pp, low confidence |
| LiveCodeBench Pro | 77.7% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (Text) | 48.6% | estimated ± 4.5 pp, low confidence |
| MMLU-Pro (Arcee) | 86.2% | estimated ± 3.4 pp, medium confidence |
| React Native Evals | 97.4% | estimated ± 2.4 pp, low confidence |
| ResearchClawBench | 18.0% | estimated ± 1.8 pp, medium confidence |
| SWE-Rebench | 61.7% | estimated ± 3.6 pp, medium confidence |
| ARC-AGI-3 | 0.0% | estimated ± 2.1 pp, medium confidence |
| BioMysteryBench (human-solvable) | 88.3% | estimated ± 0.6 pp, medium confidence |
| DRACO | 80.8% | estimated ± 3.0 pp, medium confidence |
| FrontierCode 1.1 Extended | 46.5% | estimated ± 2.4 pp, low confidence |
| HealthBench (raw) | 47.8% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 54.7% | estimated ± 2.5 pp, low confidence |
| HealthBench Professional | 56.3% | estimated ± 1.3 pp, medium confidence |
| HealthBench Professional (raw) | 58.2% | estimated ± 5.0 pp, medium confidence |
| HLE-Verified | 35.3% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 69.6% | estimated ± 1.3 pp, low confidence |
| LiveCodeBench | 100.0% | estimated ± 8.9 pp, low confidence |
| Agents' Last Exam | 25.2% | measured |
| OpenHarmony Bench | 53.8% | measured |
| IMOAnswerBench | 88.4% | measured |
| MMMLU | 89.1% | estimated ± 1.1 pp, medium confidence |
| MMLU | 93.4% | estimated ± 1.1 pp, low confidence |