With this initiative, Microsoft is marking a significant strategic shift. Long seen as a close ally of OpenAI, into which it has invested several billion dollars, the Redmond-based company is now positioning itself as a more "neutral" player in the AI ecosystem. Satya Nadella, Microsoft's CEO, emphasized the importance of allowing developers to "mix and match" models according to their needs, while guaranteeing the same reliability as that provided for OpenAI models.

The Azure offering will include xAI's Grok 3 and Grok 3 mini models, Meta's Llama models, and those from Mistral and Black Forest Labs. In total, more than 1,900 AI models are now available to Azure customers, all hosted directly in Microsoft data centers to ensure availability even during periods of high demand.

A new agent to automate software development

Another major announcement: Microsoft has unveiled an advanced version of GitHub Copilot, now referred to as a "coding agent." Unlike previous versions, which assisted developers by generating contextual code, this new agent is capable of autonomously performing development tasks based on general instructions.

For example, a user can submit a bug description and a fix strategy, and the agent will execute the entire task before submitting its work for validation. This feature is positioned against Codex, the new competing agent recently unveiled by OpenAI.

Customized AI agents for businesses

Microsoft has also unveiled Azure Foundry, a platform that allows companies to create their own internal AI agents tailored to specific business use cases. These agents can be based on one or more models available in Azure, offering great flexibility for integration.

According to Asha Sharma, vice president of product for Microsoft AI Platforms, the diversity of models will allow businesses to choose the solution that best fits their needs, while leveraging Azure's ability to reserve and share resources across the most popular models.

Microsoft has also introduced a symbolic but fundamental new feature: AI agents will now have a digital ID equivalent to that of human employees in corporate information systems. This is a way of anchoring these entities more deeply in the digital organization of companies.

For Bob O'Donnell, president of TECHnalysis Research, this approach represents a major step forward and a turning point: "Treating agents as digital coworkers opens up new capabilities, while raising new questions about the impact of AI in the workplace."

With these announcements, the Redmond-based company is continuing to build a broader AI ecosystem that is less dependent on a single partner and capable of attracting business customers looking for stability, diversity of choice, and reliable computing power.