DeepSeek V4 Pro 0813 places in the top three on NL2Repo-Bench.
DeepSeek Aug 2026 Open weights Flagship
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Humanity's Last Exam | 60% | #12 of 147 | Model card | |
| SimpleQA Verified | 52.9% | #24 of 84 | Epoch AI | |
| BullshitBench | 32.7% | #46 of 107 | Official leaderboard | |
| BullshitBench V2 | 35% | #64 of 122 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | #24 of 202 | Epoch AI | |
| GPQA Diamond | 91.7% | #28 of 219 | Epoch AI | |
| DTBench | 89.8% | #30 of 178 | Official leaderboard | |
| CritPt | 18% | #35 of 160 | Epoch AI | |
| ProofBench | 49% | #14 of 59 | Official leaderboard | |
| ARC-AGI-2 | 61.3% | #32 of 95 | Official leaderboard | |
| ARC-AGI | 90.5% | #32 of 92 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 64.6% | #39 of 89 | Epoch AI | |
| FrontierMath Tier 4 v2 | 26.8% | #41 of 71 | Epoch AI | |
| ProofBench v1.1 | 50% | #30 of 51 | Official leaderboard | |
| Prophet Arena | 84.4 | #16 of 18 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| NL2Repo-Bench | 61.5% | #3 of 28 | Model card | |
| Vibe Code Bench | 82.3% | #24 of 115 | Official leaderboard | |
| WeirdML | 66.2% | #24 of 112 | Official leaderboard | |
| ALE-Bench | 1,403 | #27 of 121 | Epoch AI | |
| WebDev Arena | 1,581 | #24 of 105 | Epoch AI | |
| DSBench (DeepSeek internal) | 71.1% | #1 of 3 | Model card | |
| Code Migration | 41.5% | #29 of 78 | Official leaderboard | |
| Code Migration v1.1 | 52% | #14 of 29 | Official leaderboard | |
| IOI v2 (Vals) | 51.6% | #26 of 45 | Official leaderboard | |
| DeepSWE | 62.7% | #31 of 51 | Model card | |
| Vibe Code Bench 1-100 | 17.5% | #16 of 26 | Official leaderboard | |
| MirrorCode | 33.3% | #20 of 28 | Epoch AI | |
| FrontierCode | 28.6% | #33 of 46 | Epoch AI |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Toolathlon | 74.1% | #5 of 37 | Model card | |
| Terminal-Bench 2.1 | 87.9% | #10 of 53 | Model card | |
| Mystery Game Puzzles | 43% | #16 of 80 | Epoch AI | |
| Toolathlon-Verified | 74.4% | #6 of 25 | Official leaderboard | |
| Chess Puzzles | 46% | #25 of 83 | Epoch AI | |
| AutomationBench | 31.8% | #25 of 41 | Model card | |
| WitnessBench | 5.50% | #8 of 12 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CorpFin v2 | 65.4% | #29 of 119 | Official leaderboard | |
| Legal Research Bench | 40.9% | #30 of 88 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 7.50% | #28 of 80 | Official leaderboard | |
| TaxEval v2 | 73.1% | #53 of 129 | Official leaderboard | |
| Tax Agent Bench | 26.8% | #36 of 79 | Official leaderboard | |
| MedCode | 42.5% | #48 of 96 | Official leaderboard | |
| PolicyBench | 83% | #25 of 50 | Official leaderboard | |
| LegalBench | 82.4% | #71 of 138 | Official leaderboard | |
| Vals Legal Index | 69% | #17 of 33 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 10.8% | #18 of 33 | Official leaderboard | |
| Vals Finance Agent v2 | 50.4% | #44 of 80 | Official leaderboard | |
| MedScribe | 80.2% | #55 of 98 | Official leaderboard | |
| APEX-Agents | 47.3% | #35 of 56 | Epoch AI | |
| Excel Modeling Benchmark | 52.8% | #53 of 76 | Official leaderboard | |
| Vals Index v2.2 | 53% | #23 of 32 | Official leaderboard | |
| Vals Finance Agent v2.1 | 57.1% | #27 of 35 | Official leaderboard | |
| Agents' Last Exam | 25.7% | #20 of 25 | Model card | |
| Excel Modeling Benchmark v1.1 | 65.1% | #30 of 37 | Official leaderboard | |
| Vals Finance Index | 52.9% | #29 of 35 | Official leaderboard | |
| Immigration Bench | 61.7% | #22 of 24 | Official leaderboard | |
| Thrive Holdings helpdesk ticket resolution benchmark | 32.7% | #14 of 15 | Official leaderboard | |
| Holdings-IndividualTaxBench | 64.4% | #13 of 13 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| VoxelBench | 1,584 | #35 of 56 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CyberBench v1.1 (Vals) | 72.9% | #12 of 43 | Official leaderboard | |
| CyberGym | 83.3% | #10 of 33 | Model card | |
| SRE-Bench | 1.91% | #23 of 33 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 59.3 | #37 of 149 | Official leaderboard | |
| LMCA | 45.5% | #37 of 139 | Epoch AI | |
| SuperCLUE | 70.1 | #6 of 18 | Official leaderboard | |
| ACCoRD | 70.2 | #51 of 150 | Official leaderboard |
DeepSeek's models in color, against every other model on the Epoch Capabilities Index, by release date.
The benchmarks the lab chose to headline, with the scores it gave.