DeepSeek · model
DeepSeek V3 benchmark scores
As of 2026-10-07, DeepSeek V3 (DeepSeek) has measured scores on 18 benchmarks and estimated scores on 75 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 55.7% | measured |
| AA-HLE | 2.9% | measured |
| AA-SciCode | 35.8% | measured |
| AIME26 | 95.4% | estimated ± 0.6 pp, low confidence |
| Artificial Analysis Intelligence Index | 8.5% | measured |
| CritPt | 0.0% | measured |
| GDPval-AA | 0.0% | measured |
| GPQA Diamond | 59.1% | estimated ± 0.4 pp, high confidence |
| HLE | 7.7% | estimated ± 8.2 pp, low confidence |
| IFBench | 55.5% | estimated ± 6.9 pp, low confidence |
| AA-LCR | 29.3% | measured |
| LiveCodeBench v6 | 71.9% | estimated ± 3.1 pp, high confidence |
| AA-Omniscience Accuracy | 25.5% | measured |
| SciCode | 36.0% | estimated ± 3.2 pp, high confidence |
| BrowseComp | 7.7% | estimated ± 9.0 pp, low confidence |
| HLE w/ tools | 31.6% | estimated ± 6.1 pp, low confidence |
| IFEval | 86.1% | measured |
| LongBench v2 | 0.0% | estimated ± 5.9 pp, low confidence |
| τ²-bench results | 22.8% | measured |
| AA Coding Index | 23.0% | measured |
| APEX-Agents | 0.0% | estimated ± 10.0 pp, low confidence |
| APEX-Agents-AA | 4.7% | estimated ± 9.1 pp, medium confidence |
| BioMysteryBench (human-difficult) | 31.7% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 17.0% | estimated ± 4.5 pp, medium confidence |
| SWE-bench Verified | 42.0% | measured |
| AutomationBench | 28.4% | estimated ± 9.7 pp, low confidence |
| BFCL v4 | 53.8% | estimated ± 2.6 pp, high confidence |
| CyberGym | 46.2% | estimated ± 5.3 pp, low confidence |
| JobBench | 0.0% | estimated ± 8.6 pp, low confidence |
| SWE-bench Pro | 17.9% | estimated ± 3.2 pp, medium confidence |
| WideResearch | 41.6% | estimated ± 9.1 pp, low confidence |
| AA Agentic Index | 0.8% | measured |
| MMLU-Pro | 75.9% | measured |
| FrontierMath v2 (Tier 4) | 1.2% | estimated ± 4.5 pp, medium confidence |
| FrontierMath v2 (Tiers 1-3) | 1.7% | measured |
| GPQA | 59.1% | measured |
| AA-IFBench | 34.8% | measured |
| Gert Labs | 32.5% | estimated ± 8.5 pp, low confidence |
| AA-AnalystAgent | 0.0% | estimated ± 12.0 pp, low confidence |
| AA EnterpriseOps-Gym | 1.2% | estimated ± 3.6 pp, low confidence |
| AA Harvey LAB | 79.4% | estimated ± 1.2 pp, low confidence |
| ARC-AGI-1 | 1.0% | estimated ± 5.4 pp, low confidence |
| ARC-AGI-2 | 0.0% | estimated ± 11.6 pp, low confidence |
| CursorBench 3.1 | 9.0% | estimated ± 3.4 pp, low confidence |
| CursorBench 3.2 | 13.4% | estimated ± 4.4 pp, medium confidence |
| FrontierCode 1.1 Main | 0.0% | estimated ± 2.7 pp, low confidence |
| FrontierSWE v2 | 0.0% | estimated ± 14.6 pp, low confidence |
| OSWorld-Verified | 55.6% | estimated ± 8.7 pp, low confidence |
| Vals GPQA Diamond | 41.9% | estimated ± 4.8 pp, medium confidence |
| Vals LiveCodeBench | 8.9% | estimated ± 9.9 pp, low confidence |
| Vals MMLU-Pro | 2.8% | estimated ± 1.2 pp, medium confidence |
| Vals SWE-bench | 50.4% | estimated ± 5.4 pp, low confidence |
| AA AutomationBench | 3.7% | estimated ± 7.0 pp, low confidence |
| GDP.pdf | 9.4% | estimated ± 5.4 pp, low confidence |
| AA ITBench | 99.9% | estimated ± 3.6 pp, low confidence |
| AA Tau3 Banking | 22.0% | estimated ± 5.8 pp, low confidence |
| CursorBench 4.0 | 5.2% | estimated ± 1.6 pp, medium confidence |
| CWE-bench v1 | 0.0% | estimated ± 6.9 pp, low confidence |
| DeepSWE | 0.2% | estimated ± 6.4 pp, low confidence |
| HLE w/o tools | 6.2% | estimated ± 4.9 pp, medium confidence |
| PostTrainBench v1.1 | 8.3% | estimated ± 6.1 pp, low confidence |
| Toolathlon-Verified | 3.2% | estimated ± 2.3 pp, low confidence |
| Vibe Code Bench | 0.6% | estimated ± 8.1 pp, low confidence |
| ExploitGym | 0.0% | estimated ± 12.2 pp, low confidence |
| C-Eval | 72.8% | estimated ± 0.9 pp, low confidence |
| Claw-Eval | 5.6% | estimated ± 13.6 pp, low confidence |
| MCP Atlas | 64.6% | estimated ± 7.8 pp, low confidence |
| MCP-Tasks | 100.0% | estimated ± 8.2 pp, low confidence |
| MMAnswerBench | 82.8% | estimated ± 1.5 pp, low confidence |
| MMLU-Redux | 87.9% | estimated ± 0.7 pp, high confidence |
| NL2Repo | 15.1% | estimated ± 5.1 pp, low confidence |
| QwenClawBench | 55.9% | estimated ± 6.0 pp, low confidence |
| SuperGPQA | 70.5% | estimated ± 7.5 pp, medium confidence |
| Toolathlon | 26.9% | estimated ± 6.3 pp, low confidence |
| HealthBench Hard | 0.0% | estimated ± 5.1 pp, low confidence |
| MMLU-Pro (Arcee) | 77.1% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 69.9% | estimated ± 3.5 pp, high confidence |
| ResearchClawBench | 1.7% | estimated ± 2.0 pp, medium confidence |
| SWE-Rebench | 57.0% | estimated ± 3.6 pp, low confidence |
| ARC-AGI-3 | 0.0% | estimated ± 2.1 pp, medium confidence |
| FrontierMath (legacy) | 2.8% | estimated ± 0.7 pp, low confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| DRACO | 61.5% | estimated ± 3.0 pp, low confidence |
| FrontierCode 1.1 Extended | 23.0% | estimated ± 3.4 pp, low confidence |
| HealthBench (raw) | 4.6% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 53.9% | estimated ± 3.7 pp, low confidence |
| HealthBench Professional | 6.5% | estimated ± 4.8 pp, low confidence |
| HealthBench Professional (raw) | 19.1% | estimated ± 5.0 pp, low confidence |
| HLE-Verified | 15.3% | estimated ± 2.0 pp, low confidence |
| LABBench2 | 36.4% | estimated ± 1.8 pp, low confidence |
| LiveCodeBench | 37.6% | measured |
| OpenHarmony Bench | 51.0% | estimated ± 2.4 pp, medium confidence |
| MMLU | 89.1% | estimated ± 1.1 pp, medium confidence |