KUALA LUMPUR – Microsoft has announced that Deepseek’s reasoning model, R1, is now available on its Azure AI Foundry service, a platform that consolidates various artificial intelligence (AI) services for enterprises.
In a blog post on Wednesday, Microsoft stated that the version of R1 available on Azure AI Foundry has “undergone rigorous red teaming and safety evaluations”, which include “automated assessments of model behaviour and extensive security reviews to mitigate potential risks”.
Microsoft also revealed that in the near future, customers will have access to “distilled” versions of R1, enabling them to run the model locally on Copilot+ PCs.
“As we continue expanding the model catalog in Azure AI Foundry, we’re excited to see how developers and enterprises leverage… R1 to tackle real-world challenges and deliver transformative experiences,” the company said.
Deepseek R1 joins a growing catalogue of more than 1,800 models on Azure AI Foundry, encompassing frontier, open-source, industry-specific, and task-based AI models.
The integration aims to provide businesses with a scalable and enterprise-ready AI platform that meets service-level agreements (SLAs), security standards, and responsible AI commitments.
According to Microsoft, one of the key advantages of using Deepseek R1 on Azure AI Foundry is the speed at which developers can experiment, iterate, and integrate AI into their workflows. The platform includes built-in model evaluation tools that allow users to compare outputs, benchmark performance, and scale AI-powered applications.
To ensure safety and compliance, Microsoft highlighted that Deepseek R1 has undergone extensive security reviews, including automated behaviour assessments.
Azure AI Foundry also incorporates content safety features, such as built-in content filtering with opt-out options, and a Safety Evaluation System to allow customers to test applications before deployment.
DeepSeek R1 is now accessible through a serverless endpoint in the model catalogue on Azure AI Foundry. Users can deploy the model, obtain inference API keys, and access a testing playground via the platform. Additionally, Microsoft has made further resources available on GitHub, including step-by-step integration guides.
Microsoft further announced that customers will soon be able to use “distilled” versions of Deepseek R1 on Copilot+ PCs, expanding the model’s accessibility and flexibility for local deployments. – January 30, 2025