Microsoft added support for the recently released DeepSeek-R1 artificial intelligence (AI) model on Wednesday on Azure AI Foundry and Github. Inference-centric AI models are now accessible through these platforms. The Redmond-based tech giant also stressed that it has conducted rigorous security assessments to ensure that the model is safe for personal and commercial purposes. Additionally, the company brings the distilled version of the R1 model to Copilot+ PC users through an AI toolkit.
Microsoft brings DeepSeek-R1 to Azure AI Foundry, GitHub
The tech giant announced in a blog post that the DeepSeek-R1 AI model is now available in the model catalogs of Azure AI Foundry and Github. It is worth noting that Azure AI Foundry is an enterprise-centric platform where developers can build, evaluate and develop generated AI applications and custom co-pilots.
Announced the news, Asha Sharma, vice president of AI Platforms at Microsoft, said: “One of the key advantages of using DeepSeek R1 or any other model on Azure AI Foundry is that developers can try, iterate and integrate AI into their AI efforts. process.”
The post also highlighted that the company conducted a red team and security assessment of the R1 model, including an automatic assessment of model behavior and an extensive security review. Microsoft said it has ensured potential risks mitigated. Azure AI Foundry also provides additional security through its content security filtering system and security assessment system.
To use DeepSeek-R1, users can search for it in the model directory. Once it is discovered, they will have to open the model card and click Deploy. This will provide the user with the inference application programming interface (API) required to run the model.
Distled DeepSeek-R1 comes to Copilot+ PCS user
In another article, Microsoft announced that it would bring an optimized version of DeepSeek-R1 to Copilot+ PC. These distilled versions will first arrive at the Snapdragon X power device and then add it to the Intel Core Ultra 200V chipset and more. The first model version will be the DeepSeek-R1-Distill-Qwen-1.5b model that can be accessed through the AI toolkit. Variants of 7b and 14b will be added later.
