Azure Updates: DeepSeek R1; o3-mini model; Azure DDoS

Microsoft CVP Asha Sharma announced that DeepSeek R1 is now available for Azure AI Foundry and GitHub. DeepSeek R1 was added to the model catalog, joining 1, 800 other models currently available to customers like task-based and industry-specific AI models. DeepSeek R1 is intended to provide integrated AI while being able to meet security, responsible AI, and SLA standards. 

According to Sharma, one of the key advantages with DeepSeek R1 is the ability to rapidly build, experiment, and iterate with AI integrated into workflows. Azure customers can search for DeepSeek R1 in the model catalog, deploy it with an inference API and key, then deploy with various clients. 

VP of Product Yina Arenas announced the availability of the o3-mini reasoning model for Azure OpenAI Service. The new model is intended to add cost efficiencies compared with o1-mini, supporting enhanced reasoning, effort control, and lower latency. New features include a reasoning effort parameter, structured JSON Schema outputs, functions and tools support, developer and system message compatibility, as well as math and science reasoning. 

Senior PM Manager Amir Dahan highlighted Azure DDoS in action during the 2024 Christmas shopping season. Cyber criminals launched sophisticated DDoS-for-hire operations, botnets, and worked to bypass CDN protections. Daily attack volume soared to 3800 attacks daily, including highly volumetric attacks over 10M pps.