DeepSeek, a Chinese AI startup, has launched its first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1, with significant performance in reasoning tasks. While DeepSeek-R1-Zero faced challenges like repetition and language mixing, DeepSeek-R1 improved upon these issues by incorporating cold-start data before reinforcement learning, achieving performance on par with OpenAI-o1 in math, code, and reasoning tasks. To support the research community, DeepSeek has open-sourced these models and six dense models distilled from DeepSeek-R1, with DeepSeek-R1-Distill-Qwen-32B surpassing OpenAI-o1-mini in benchmarks. A tutorial outlines deploying DeepSeek models on Cerebrium's serverless architecture, highlighting cost efficiency, security, ease of deployment, and scalability. By using Cerebrium, users can create scalable, OpenAI-compatible endpoints with vLLM, leveraging streamlined infrastructure and security compliance to manage AI models effectively.