Microsoft Unveils General Availability of Azure AI with Copilot and Integration of Meta Llama 4
News
Microsoft Launches Azure AI with Copilot and Meta Llama 4 Integration
Microsoft has officially announced the general availability (GA) of Copilot in Azure, along with integrating Meta’s latest Llama 4 models into Azure AI Foundry and Azure Databricks.
Microsoft’s Copilot Reaches General Availability
Microsoft confirmed that Copilot in Azure has transitioned from public preview to general availability. Launched in May 2024, Copilot has quickly gained traction among a variety of organizations. Ruhiyyih Mahalati from Microsoft noted, “Since launching Copilot in an ungated public preview, we have facilitated millions of prompts for hundreds of thousands of users,” indicating its widespread adoption. Notably, it is estimated that Copilot is saving Microsoft over 30,000 developer hours each month.
New Features and Improvements
With the GA release of Copilot, users can expect several enhancements, including:
- Improved response times, with enhancements to both frontend and backend systems leading to over a 30% increase.
- An updated user interface designed with high accessibility standards in mind.
- A commitment to 99.9% uptime for Copilot services.
- Robust testing protocols aligned with Microsoft’s Responsible AI principles to mitigate harmful behavior.
- Support for 19 languages, expanding its usability across different markets.
- New functionalities such as Terraform configuration authoring and diagnostics for Azure Kubernetes Service.
Additionally, Copilot is now accessible on the Azure Mobile App, introducing features like real-time AI chat streaming, improved entry points, cost management skills, and enhanced accessibility and localization options.
Meta’s Llama 4 Models Now Available
Microsoft also introduced Meta’s new Llama 4 models—Scout and Maverick—on Azure AI Foundry and Azure Databricks. These models are engineered to tackle complex tasks by combining text and visual data in a single model framework.
- Llama 4 Scout: This model is specifically designed for tasks such as summarization, personalization, and long-context reasoning. It can handle up to 10 million tokens and operates efficiently on a single H100 GPU.
- Llama 4 Maverick: Featuring 17 billion active parameters and a Mixture of Experts (MoE) architecture, Maverick is tailored for multilingual and multimodal chat applications.
The Llama 4 models incorporate essential safety features throughout their development phases and are aligned with Azure’s stringent security and compliance standards. The ability to blend different types of input seamlessly, alongside the scalability of MoE, makes the Llama 4 family an ideal choice for enterprises looking to develop sophisticated AI solutions while maintaining performance and cost-effectiveness.
