DeepSeek·released 2026-08-131 source
DeepSeek V4 Pro 0813 benchmark scores: 15 benchmarks tracked. 33% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 15 benchmarks·best result 98.61 on OTIS Mock AIME 2024-2025 (reproduced)·5 of 15 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.61accuracy (%) | reproduced· optimizedT1 | 2026-08-13Epoch AI |
| ARC-AGI | 90.5accuracy (%) | unverified· optimizedT1 | 2026-08-13https://arcprize.org/leaderboard |
| DTBench | 89.78accuracy (%) | unverifiedT1 | 2026-08-13https://conceptualreasoning.ai/dtbench |
| GPQA diamond | 88.89accuracy (%) | reproduced· optimizedT1 | 2026-08-13Epoch AI |
| WeirdML | 66.22accuracy (%) | unverifiedT1 | 2026-08-13https://htihle.github.io/weirdml.html |
| FrontierMath-Tiers-1-3-v2-Private | 64.56accuracy (%) | reproducedT1 | 2026-08-13Epoch AI |
| ARC-AGI-2 | 61.25accuracy (%) | unverifiedT2 | 2026-08-13 |
| LMCA | 53.47accuracy (%) | unverifiedT1 | 2026-08-13https://conceptualreasoning.ai/lmca |
| SimpleQA Verified | 52.91accuracy (%) | unverifiedT2 | 2026-08-13 |
| ProofBench | 50accuracy (%) | unverifiedT1 | 2026-08-13https://www.vals.ai/benchmarks/proof_bench |
| APEX-Agents | 47.3accuracy (%) | unverifiedT2 | 2026-08-13 |
| Chess Puzzles | 43.18accuracy (%) | unverifiedT2 | 2026-08-13 |
| Mystery Game Puzzles | 37.21accuracy (%) | reproducedT1 | 2026-08-13Epoch AI |
| FrontierCode | 28.55accuracy (%) | unverifiedT1 | 2026-08-13https://cognition.com/frontiercode |
| FrontierMath-Tier-4-v2-Private | 26.83accuracy (%) | reproducedT1 | 2026-08-13Epoch AI |
Claims drawn from cited facts, not live model generation.
This model was developed by DeepSeek and released in 2026.
2 cited facts
This model has scores on 15 tracked benchmarks and currently leads 0 of them. The record’s evidence status is independently reproduced.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 28.22 points.
1 cited fact
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
DeepSeek V4 Pro 0813 is an AI model developed by DeepSeek, released 2026-08-13. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
DeepSeek V4 Pro 0813 has recorded scores on 15 benchmarks, each shown with its evidence status.
DeepSeek V4 Pro 0813 has recorded scores on 15 benchmarks — GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, LMCA, SimpleQA Verified, and 9 more. The full table above shows each score with its evidence status.
5 of 15 recorded scores (33%) are independently reproduced rather than self-reported by the lab.
DeepSeek V4 Pro 0813 has 15 tracked claims: 5 independently reproduced, 10 unverified.