DeepSeek Benchmarks
Published scores for DeepSeek models next to contemporary models. Every number is copied from official model cards — we did not run these.
Why that distinction matters. Running benchmarks
yourself needs API budget, a fixed evaluation environment and correct configuration for each
task. What we do is compile the numbers vendors publish, with sources attached.
The difference is real: vendor-reported scores can be selectively published. But it is also the only way to keep this page fresh at near-zero cost. If demand justifies it, we will run them ourselves later.
The difference is real: vendor-reported scores can be selectively published. But it is also the only way to keep this page fresh at near-zero cost. If demand justifies it, we will run them ourselves later.
Benchmarks
46
Models
23
DeepSeek models
5
Refresh
weekly
How to read it.
Blue cells are DeepSeek's own models; bold is the row
maximum. Evaluation settings differ between tables (shots, output-length caps), so
compare across tables with care.
Other
| Benchmark | Setting | DeepSeek-V2 | Qwen2.5 72B | LLaMA3.1 405B | DeepSeek-V3 | Claude-3.5-Sonnet-1022 | GPT-4o 0513 | DeepSeek V3 | DeepSeek V2.5-0905 | Qwen2.5 72B-Inst. | Llama3.1 405B-Inst. | OpenAI o1-mini | DeepSeek R1 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AGIEval (Acc.) | 0-shot | 57.5 | 75.8 | 60.6 | 79.6 | — | — | — | — | — | — | — | — |
| AIME 2024 (Pass@1) | — | — | — | — | 16 | 9.3 | 39.2 | 16.7 | 23.3 | 23.3 | 63.6 | 79.8 | |
| ARC-Challenge (Acc.) | 25-shot | 92.2 | 94.5 | 95.3 | 95.3 | — | — | — | — | — | — | — | — |
| ARC-Easy (Acc.) | 25-shot | 97.6 | 98.4 | 98.4 | 98.9 | — | — | — | — | — | — | — | — |
| Aider-Edit (Acc.) | — | — | — | — | 84.2 | 72.9 | 79.7 | 71.6 | 65.4 | 63.9 | — | — | |
| Aider-Polyglot (Acc.) | — | — | — | — | 45.3 | 16 | 49.6 | 18.2 | 7.6 | 5.8 | 32.9 | 53.3 | |
| AlpacaEval2.0 (LC-winrate) | — | — | — | — | 52 | 51.1 | 70 | — | — | — | 57.8 | 87.6 | |
| ArenaHard (GPT-4-1106) | — | — | — | — | 85.2 | 80.4 | 85.5 | — | — | — | 92 | 92.3 | |
| C-Eval (Acc.) | 5-shot | 81.4 | 89.2 | 72.5 | 90.1 | 76.7 | 76 | 86.5 | 79.5 | 86.1 | 61.5 | 68.9 | 91.8 |
| C-SimpleQA (Correct) | — | — | — | — | 55.4 | 59.3 | 68 | 54.1 | 48.4 | 50.4 | 40.3 | 63.7 | |
| C3 (Acc.) | 0-shot | 77.4 | 76.7 | 79.7 | 78.6 | — | — | — | — | — | — | — | — |
| CCPM (Acc.) | 0-shot | 93 | 88.5 | 78.6 | 92 | — | — | — | — | — | — | — | — |
| CLUEWSC (EM) | 5-shot | 82 | 82.5 | 83 | 82.7 | 85.4 | 87.9 | 90.9 | 90.4 | 91.4 | 84.7 | 89.9 | 92.8 |
| CMMLU (Acc.) | 5-shot | 84 | 89.5 | 73.7 | 88.8 | — | — | — | — | — | — | — | — |
| CMRC (EM) | 1-shot | 77.4 | 75.8 | 76 | 76.3 | — | — | — | — | — | — | — | — |
| CNMO 2024 (Pass@1) | — | — | — | — | 13.1 | 10.8 | 43.2 | 10.8 | 15.9 | 6.8 | 67.6 | 78.8 | |
| CRUXEval-I (Acc.) | 2-shot | 52.5 | 59.1 | 58.5 | 67.3 | — | — | — | — | — | — | — | — |
| CRUXEval-O (Acc.) | 2-shot | 49.8 | 59.9 | 59.9 | 69.8 | — | — | — | — | — | — | — | — |
| Codeforces (Percentile) | — | — | — | — | 717 | 759 | 1134 | 35.6 | 24.8 | 25.3 | 1820 | 2029 | |
| DROP (F1) | 3-shot | 80.4 | 80.6 | 86 | 89 | 88.3 | 83.7 | 91.6 | 87.8 | 76.7 | 88.7 | 83.9 | 92.2 |
| FRAMES (Acc.) | — | — | — | — | 72.5 | 80.5 | 73.3 | 65.4 | 69.8 | 70 | 76.9 | 82.5 | |
| GPQA-Diamond (Pass@1) | — | — | — | — | 65 | 49.9 | 59.1 | 41.3 | 49 | 51.1 | 60 | 71.5 | |
| GSM8K (EM) | 8-shot | 81.6 | 88.3 | 83.5 | 89.3 | — | — | — | — | — | — | — | — |
| HellaSwag (Acc.) | 10-shot | 87.1 | 84.8 | 89.2 | 88.9 | — | — | — | — | — | — | — | — |
| HumanEval (Pass@1) | 0-shot | 43.3 | 53 | 54.9 | 65.2 | — | — | — | — | — | — | — | — |
| HumanEval-Mul (Pass@1) | — | — | — | — | 81.7 | 80.5 | 82.6 | 77.4 | 77.3 | 77.2 | — | — | |
| IF-Eval (Prompt Strict) | — | — | — | — | 86.5 | 84.3 | 86.1 | 80.6 | 84.1 | 86 | 84.8 | 83.3 | |
| LiveCodeBench (Pass@1-COT) | — | — | — | — | 36.3 | 34.2 | 40.5 | 29.2 | 31.1 | 30.1 | 53.8 | 65.9 | |
| LiveCodeBench-Base (Pass@1) | 3-shot | 11.6 | 12.9 | 15.5 | 19.4 | — | — | — | — | — | — | — | — |
| LongBench v2 (Acc.) | — | — | — | — | 41 | 48.1 | 48.7 | 35.4 | 39.4 | 36.1 | — | — | |
| MATH (EM) | 4-shot | 43.4 | 54.4 | 49 | 61.6 | — | — | — | — | — | — | — | — |
| MATH-500 (EM) | — | — | — | — | 78.3 | 74.6 | 90.2 | 74.7 | 80 | 73.8 | 90 | 97.3 | |
| MBPP (Pass@1) | 3-shot | 65 | 72.6 | 68.4 | 75.4 | — | — | — | — | — | — | — | — |
| MMLU (Acc.) | 5-shot | 78.4 | 85 | 84.4 | 87.1 | 88.3 | 87.2 | 88.5 | 80.6 | 85.3 | 88.6 | 85.2 | 90.8 |
| MMLU-Pro (Acc.) | 5-shot | 51.4 | 58.3 | 52.8 | 64.4 | 78 | 72.6 | 75.9 | 66.2 | 71.6 | 73.3 | 80.3 | 84 |
| MMLU-Redux (Acc.) | 5-shot | 75.6 | 83.2 | 81.3 | 86.2 | 88.9 | 88 | 89.1 | 80.3 | 85.6 | 86.2 | 86.7 | 92.9 |
| MMMLU-non-English (Acc.) | 5-shot | 64 | 74.8 | 73.8 | 79.4 | — | — | — | — | — | — | — | — |
| Math | — | — | — | — | — | — | — | — | — | — | — | — | |
| NaturalQuestions (EM) | 5-shot | 38.6 | 33.2 | 41.5 | 40 | — | — | — | — | — | — | — | — |
| PIQA (Acc.) | 0-shot | 83.9 | 82.6 | 85.9 | 84.7 | — | — | — | — | — | — | — | — |
| RACE-High (Acc.) | 5-shot | 52.6 | 50.3 | 56.8 | 51.3 | — | — | — | — | — | — | — | — |
| RACE-Middle (Acc.) | 5-shot | 73.1 | 68.1 | 74.2 | 67.1 | — | — | — | — | — | — | — | — |
| SWE Verified (Resolved) | — | — | — | — | 50.8 | 38.8 | 42 | 22.6 | 23.8 | 24.5 | 41.6 | 49.2 | |
| SimpleQA (Correct) | — | — | — | — | 28.4 | 38.2 | 24.9 | 10.2 | 9.1 | 17.1 | 7 | 30.1 | |
| TriviaQA (EM) | 5-shot | 80 | 71.9 | 82.7 | 82.9 | — | — | — | — | — | — | — | — |
| WinoGrande (Acc.) | 5-shot | 86.3 | 82.3 | 85.2 | 84.9 | — | — | — | — | — | — | — | — |
Source and caveats
| Item | Detail |
|---|---|
| Source | 官方模型卡 README（HuggingFace: deepseek-ai/*） |
| Collected | 2026-11-05 |
| Refresh | Weekly |
| Run by us | No — copied from official model cards |
| Known issues | Evaluation settings may differ between tables; vendors may publish only favourable results |
分数抄自官方模型卡，不是我们自己跑的。不同表格的评测设置可能不同，跨表比较需谨慎。