DeepSeek API兼容OpenAI/Anthropic格式，只需修改配置即可使用其SDK或兼容软件访问。需申请API密钥，并指定模型名称（部分旧模型将于2026年7月24日停用）。提供OpenAI格式的调用示例，支持流式和非流式响应，Anthropic格式调用参考相关文档。

标题：首次调用DeepSeek API | DeepSeek API文档

DeepSeek API采用与OpenAI/Anthropic兼容的接口格式。通过调整配置，用户可直接使用OpenAI/Anthropic的SDK或兼容工具调用DeepSeek API。

关键参数配置

| 参数 | 值 | |---------------|-------------------------------------------------------------------| | baseurl (OpenAI) | https://api.deepseek.com | | baseurl (Anthropic) | https://api.deepseek.com/anthropic | | api_key | 需申请API密钥 | | model* | deepseek-v4-flash、deepseek-v4-pro、即将废弃的deepseek-chat（2026/07/24停用）及deepseek-reasoner（同前） |

*注：deepseek-chat和deepseek-reasoner将分别对应deepseek-v4-flash的非思考模式与思考模式，未来建议迁移至新模型。

调用聊天API

获取API密钥后，可通过OpenAI格式的示例代码调用模型。以下为非流式请求示例（设置stream: true可启用流式响应）：

cURL示例
bash curl https://api.deepseek.com/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer ${DEEPSEEK_API_KEY}" \ -d '{ "model": "deepseek-v4-pro", "messages": [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Hello!"} ], "thinking": {"type": "enabled"}, "reasoning_effort": "high", "stream": false }'

如需Anthropic格式的调用方式，请参阅Anthropic API指南。

（注：原文中的Python和Node.js代码示例因与核心说明关联较弱，此处略去以保持简洁。）

总结评论内容：

1. 模型发布与性能表现

   • 评论指出DeepSeek-V4-Pro已发布，性能达到前沿水平，成本更低。
     "Model was released and it's amazing. Frontier level (better than Opus 4.6) at a fraction of the cost." (nthypes)
     "MMLU-Pro: Gemini-3.1-Pro at 91.0, Opus-4.6 at 89.1, GPT-5.4, Kimi2.6, and DS-V4-Pro tied at 87.5." (Aliabid94)

2. 技术细节与优化

   • 评论提到模型采用了创新的神经网络架构优化（如mHC和混合注意力机制）。
     "One of the key points is the optimization with the residual design... manifold-constrained hyper-connections (mHC)." (rvz)
     "KV cache is said to take 10% as much space as V3." (zargon)

3. 本地运行与量化需求

   • 用户关注模型在本地设备（如MacBook）上的运行可能性和量化版本的需求。
     "How long does it usually take for folks to make smaller distills of these models?" (ls612)
     "Is there a Quantized version of this?" (namegulf)

4. 多模态与功能缺失

   • 部分用户对模型缺乏原生多模态支持表示遗憾。
     "Excited that the long awaited v4 is finally out. But feel sad that it's not multimodal native." (jdeng)

5. 开源与中国背景

   • 评论提到模型的开源性质及其中国背景，部分用户持积极态度。
     "Truly open source coming from China. This is heartwarming." (sidcool)

6. 行业动态与用户疲劳

   • 用户对AI领域快速更新感到疲劳，并希望有工具帮助跟进。
     "I could really use a support group for people burnt out from trying to keep up with everything." (gbnwl)
     "At this point 'frontier model release' is a monthly cadence." (jessepcc)

7. 价格与可用性

   • 评论提到模型已在OpenRouter上线，价格相对较低。
     "Pro version is $1.74/m/input, $3.48/m/output, while flash $0.14/m/input, 0.28/m/output." (yanis_t)

8. 开发者文档先行

   • 用户赞赏开发者文档先于正式发布的做法。
     "There's something heartwarming about the developer docs being released before the flashy press release." (fblp)

9. 短期热度与长期关注

   • 有评论预测模型短期内会引发热议，但可能很快被遗忘。
     "DeepSeek4 floods every AI-related online space... Then a few weeks later it'll be forgotten by most." (raincol)

10. 基准测试与验证

    • 用户建议等待独立验证，而非完全信任官方基准测试。
      "I wouldn’t trust the benchmarks directly, but would wait for others to try it." (rvz)

11. 硬件需求与成本

    • 用户讨论在自有硬件上运行模型的可行性及成本。
      "How can you reasonably try to get near frontier on hardware you own? Maybe under 5k in cost?" (aliljet)

12. 模型比较与偏好

    • 有用户对比不同版本（如Flash与Pro）的输出效果，表达个人偏好。
      "I like the pelican I got out of deepseek-v4-flash more than the one from deepseek-v4-pro." (simonw)

总结：

评论普遍认可DeepSeek-V4-Pro的性能和成本优势，同时关注技术细节、本地运行可能性和行业影响。部分用户对多模态缺失和快速迭代的疲劳表示担忧，而开源背景和开发者文档先行获得积极评价。价格和量化版本的需求也是讨论焦点。