DeepSeek
DeepSeek-V3 (Mar 2025) benchmark results
6 results on 6 benchmarks from 4 sources.
Also listed as: deepseek-v3-0324 · DeepSeek-V3-0324 · DeepSeek V3 0324 · SWE-Exp · Deepseek-V3-0324
All benchmark results
| Benchmarks | Score | Rank | Setting | Source type | Date |
|---|---|---|---|---|---|
| LMArena Text | 1395 | 165 | Overall, style control on; 44989 votes; 95% CI 1391–1399“deepseek-v3-0324” | Official leaderboard | Oct 8, 2026 |
| GPQA Diamond | 67.6% | 166 | Epoch AI run, accuracy; effort: Mar 2025 | Independent run | Apr 1, 2025 |
| MMLU-Pro | 81.3% | 57 | Overall accuracy; data source: Self-Reported; 671B“Deepseek-V3-0324” | Vendor-reported | Mar 11, 2026 |
| SWE-bench Verified | 42.0% | 80 | agent: SWE-Exp“DeepSeek-V3-0324” | Official leaderboard | Aug 6, 2025 |
| AIME (OTIS Mock AIME 2024–2025) | 37.8% | 183 | Epoch AI run, OTIS Mock AIME 2024–2025; effort: Mar 2025 | Independent run | Apr 1, 2025 |
| MATH Level 5 | 75.5% | 38 | Epoch AI run, accuracy; effort: Mar 2025 | Independent run | Apr 1, 2025 |
FAQ
Where do these DeepSeek-V3 (Mar 2025) scores come from?
From Epoch AI Benchmarking Hub, LMArena leaderboard dataset, MMLU-Pro leaderboard (TIGER-Lab), and SWE-bench leaderboards (swebench.com). Each row links to the original leaderboard, dataset or report, and vendor-reported numbers are labelled.