DeepSeek:API 更新日誌·· 2025-05-28精選AI 評分76
DeepSeek 將 deepseek-reasoner 升級至 DeepSeek-R1-0528,推理基準分數提升
deepseek-reasoner
AI 導讀
DeepSeek 將 deepseek-reasoner 升級至 DeepSeek-R1-0528,並公佈多項基準的 Pass@1 提升。AIME 2025 由 70.0 升至 87.5(+17.5),GPQA 由 71.5 升至 81.0(+9.5),LCB_v6 由 63.5 升至 73.3(+9.8),Aider 由 57.0 升至 71.6(+14.6)。
推薦理由
DeepSeek-R1-0528公佈多項基準提升數據,並提示複雜推理任務較舊版消耗更多 tokens,可供比較版本能力表現與用量。
正文 · 原文
deepseek-reasoner Model Upgraded to DeepSeek-R1-0528:
- Enhanced Reasoning Capabilities
- Significant benchmark improvements (Pass@1)
- AIME 2025: 70.0 → 87.5 (+17.5)
- GPQA: 71.5 → 81.0 (+9.5)
- LCB_v6: 63.5 → 73.3 (+9.8)
- Aider: 57.0 → 71.6 (+14.6)
- Note: Complex reasoning tasks may consume more tokens compared to legacy R1 version.
- Significant benchmark improvements (Pass@1)
- Optimized Front-end Development
- Generated web pages and games now feature improved aesthetics.
- Reduced Hallucinations
- Significantly suppressed hallucination issues present in legacy R1 version.
- JSON Output & Function Calling Support
- Function call performance:
- Tau-bench score: 53.5 (Airline) / 63.9 (Retail)
- Function call performance:
來源:DeepSeek:API 更新日誌 · api-docs.deepseek.com