DeepSeek V4.1 Flash 1017 ranks first on 2 of the 97 benchmarks with five or more models scored.
DeepSeek Oct 2026 Open weights Mid-size
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 | |
|---|---|---|---|---|
| MultiChallenge | 74.9% | #5 of 42 | Official leaderboard | |
| Extended NYT Connections | 92.8% | #16 of 103 | Official leaderboard | |
| Humanity's Last Exam | 42.6% | #61 of 154 | Lab model card | |
| SimpleQA Verified | 42.5% | #49 of 94 | Epoch AI run | |
| Kimi Hallu Bench | 13.1% | #5 of 6 | Official leaderboard | |
| HANDBOOK.md | 24% | #14 of 16 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 99.4% | #20 of 212 | Epoch AI run | |
| GPQA Diamond | 93.2% | #23 of 229 | Epoch AI run | |
| MathArena (Apex, IMO, USAMO) | 72.1% | #2 of 12 | Lab model card | |
| ARC-AGI-2 | 78.1% | #21 of 98 | Official leaderboard | |
| ARC-AGI | 94.5% | #24 of 95 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 71.2% | #28 of 86 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 46.3% | #25 of 68 | Official leaderboard | |
| BrokenArXiv | 48.7% | #4 of 6 | Official leaderboard | |
| Riemann-bench | 38.4% | #15 of 20 | Official leaderboard | |
| ArXivMath | 42.5% | #5 of 6 | Official leaderboard | |
| CritPt | 50% | #10 of 11 | Official leaderboard | |
| CritPt v2 | 50% | #10 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| NL2Repo-Bench | 68.2% | #1 of 30 | Lab model card | |
| DeepSWE | 77% | #4 of 56 | Lab model card | |
| Vibe Code Bench | 87.7% | #14 of 120 | Official leaderboard | |
| ALE-Bench | 2,012 | #20 of 126 | Official leaderboard | |
| Tencent WorkBuddy Bench | 78.3% | #3 of 14 | Official leaderboard | |
| APEX-SWE | 49.1% | #16 of 65 | Official leaderboard | |
| Code Migration v1.1 | 67.5% | #8 of 30 | Official leaderboard | |
| ReactBench | 34.9% | #8 of 28 | Official leaderboard | |
| Code Migration | 44.2% | #25 of 78 | Official leaderboard | |
| CursorBench | 47.7% | #9 of 21 | cursor.com, via Epoch AI | |
| CursorBench 4.0 | 47.7% | #10 of 21 | Official leaderboard | |
| WeirdML v3 | 9.10% | #16 of 27 | Official leaderboard | |
| FrontierCS 2 | 14.6% | #12 of 19 | Official leaderboard | |
| DAXBench v2 | 74.5% | #30 of 47 | Official leaderboard | |
| Android Bench 2.0 | 17.3% | #11 of 17 | Official leaderboard | |
| FrontierCode 2.0 | 20.2% | #21 of 32 | Official leaderboard | |
| Kimi Code Bench 3.0 Long-horizon | 51.5% | #5 of 6 | Official leaderboard | |
| DAYJOB: Data Science | 3.30% | #13 of 15 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LisanBench | 21107.3 | #4 of 133 | Official leaderboard | |
| AutomationBench | 60.5% | #3 of 47 | Lab model card | |
| Mystery Game Puzzles | 49% | #16 of 84 | Epoch AI run | |
| RuneBench | 5.83 | #16 of 76 | Official leaderboard | |
| Chess Puzzles | 46% | #28 of 93 | Epoch AI run | |
| DuelLab GameBench 2 | 69.9 | #19 of 57 | Official leaderboard | |
| Terminal-Bench 4.0 | 40% | #14 of 40 | Official leaderboard | |
| Terminal-Bench 4.1 | 38.5% | #17 of 43 | Official leaderboard | |
| BU Bench V2.1 | 75.0 | #7 of 16 | Official leaderboard | |
| ARC-AGI-3 | 5.81% | #12 of 27 | Official leaderboard | |
| o11y-bench | 81.3% | #11 of 22 | Official leaderboard | |
| Cresta-Bench | 32.6% | #13 of 14 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Tax Agent Bench | 44.4% | #8 of 84 | Official leaderboard | |
| APEX-Agents | 69.8% | #6 of 61 | Collected by Epoch AI, source not named | |
| APEX-Agents 1.1 | 79.4% | #9 of 65 | Official leaderboard | |
| FoodTruck Bench | $53,289 | #9 of 59 | Official leaderboard | |
| Vals Finance Agent v2 | 57% | #17 of 80 | Official leaderboard | |
| APEX-1 | 68.3% | #19 of 71 | Official leaderboard | |
| Legal Research Bench | 45.2% | #25 of 93 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 58.5% | #14 of 51 | Official leaderboard | |
| Vals Index v2.1 | 56.6% | #14 of 49 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 9.17% | #24 of 80 | Official leaderboard | |
| Vals Index v2.2 | 61.3% | #12 of 37 | Official leaderboard | |
| Vals Finance Index | 60.9% | #13 of 40 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 14.2% | #13 of 38 | Official leaderboard | |
| Vals Legal Index | 71.9% | #13 of 38 | Official leaderboard | |
| DAYJOB: Finance | 12.5% | #15 of 39 | Official leaderboard | |
| Agents' Last Exam | 35.4% | #10 of 26 | Lab model card | |
| DAYJOB: Healthcare | 6.00% | #15 of 37 | Official leaderboard | |
| Excel Modeling Benchmark | 61.8% | #32 of 76 | Official leaderboard | |
| Vals Finance Agent v2.1 | 62.9% | #18 of 40 | Official leaderboard | |
| Tuesday Work Index | 52.7 | #21 of 44 | Official leaderboard | |
| APEX-Accounting | 59.3% | #24 of 49 | Official leaderboard | |
| PRBench | 47.3% | #24 of 48 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 69.2% | #22 of 42 | Official leaderboard | |
| Agents' Last Exam v1.1 | 36.1% | #9 of 14 | Official leaderboard | |
| DAYJOB: Tax | 1.20% | #10 of 15 | Official leaderboard | |
| Hebbia Financial Services Benchmark | 60.3% | #7 of 10 | Official leaderboard | |
| DAYJOB: Accounting | 7.60% | #13 of 17 | Official leaderboard | |
| Bedside Bench | 85.7% | #14 of 17 | Official leaderboard | |
| DAYJOB: Law | 6.20% | #14 of 16 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ZeroBench | 51% | #1 of 31 | Lab model card | |
| BabyVision | 92.5% | #3 of 26 | Lab model card | |
| VoxelBench | 2,037 | #12 of 56 | Official leaderboard | |
| Chartography | 83% | #3 of 14 | Lab model card | |
| BVB | 70.0 | #13 of 45 | Official leaderboard | |
| Humanity's Sixth Sense | 30.1% | #21 of 34 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DrugDiscoveryBench | 60.6% | #5 of 20 | Official leaderboard | |
| Human Pathogen Capabilities Test | 60.5% | #23 of 86 | Official leaderboard | |
| Molecular Biology Capabilities Test | 53.9% | #25 of 90 | Official leaderboard | |
| nocot-bench (NCRI) | 81.6 | #55 of 185 | Official leaderboard | |
| DeepsecBench | 30.0 | #9 of 29 | Official leaderboard | |
| CyberOffense | 72.7% | #2 of 6 | Official leaderboard | |
| BioTIER Permit | 99.7% | #27 of 69 | Official leaderboard | |
| World-Class Bio | 49.6% | #29 of 74 | Official leaderboard | |
| CWE-bench 2 | 57.6% | #8 of 14 | Official leaderboard | |
| FORTRESS | 27.2% | #43 of 73 | Official leaderboard | |
| BioTIER Refuse | 25.9% | #50 of 69 | Official leaderboard | |
| VCT-v2 | 38.5% | #19 of 25 | Official leaderboard |
The Self-Reported Index fits one score to the 10 capability results DeepSeek reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 161.2 to 170.3. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| Terminal-Bench 4.0 | 42.3% Terminal-Bench 4.0 | 40% Leaderboard, max, Claude Code | +2.3 pts |
DeepSeek's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
The benchmarks the lab chose to headline, with the scores it gave.