DeepSeek-R1-0528 is the May 28, 2025 version of DeepSeek's reasoning model. It features advanced thinking capabilities and serves as a benchmark comparison for newer models like DeepSeek-V3.1. This model excels in complex reasoning tasks, mathematical problem-solving, and code generation through its thinking mode approach.
发布日期2025年5月28日
参数规模671B
上下文长度164K
许可证MIT
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| MMLU-Redux | language reasoning math general | 93.4 | 来源 |
| SimpleQA | general reasoning | 92.3 | 来源 |
| AIME 2024 | math reasoning | 91.4 | 来源 |
| AIME 2025 | math reasoning | 87.5 | 来源 |
| MMLU-Pro | language reasoning math general | 85.0 | 来源 |
| GPQA | reasoning general | 81.0 | 来源 |
| HMMT 2025 | math | 79.4 | 来源 |
| LiveCodeBench | reasoning general code | 73.3 | 来源 |
| Aider-Polyglot | general code | 71.6 | 来源 |
| Codeforces | math reasoning | 64.3 | 来源 |
| SWE-Bench Verified | reasoning frontend_development code | 44.6 | 来源 |
| BrowseComp-zh | reasoning search | 35.7 | 来源 |
| SWE-Bench Multilingual | reasoning code | 30.5 | 来源 |
| Humanity's Last Exam | general | 17.7 | 来源 |
| BrowseComp | reasoning search | 8.9 | 来源 |
| Terminal-Bench | reasoning code | 5.7 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| Alibaba (China) | $0.57 | $2.29 | 131K | — | — | ✓ | ✗ | ✗ |
| Cortecs | $0.59 | $2.31 | 164K | — | — | ✓ | ✗ | ✗ |
| Azure | $1.35 | $5.40 | 164K | — | — | ✓ | ✗ | ✗ |
| Azure Cognitive Services | $1.35 | $5.40 | 164K | — | — | ✓ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。