Microsoft Significantly Enhances Azure AI Studio with Llama 2 Model-as-a-Service and GPT-4 Turbo Featuring Vision Capabilities

Microsoft has emerged as a prominent player in the generative AI landscape, thanks to its early investment in OpenAI and the integration of OpenAI's technology into various services like Bing and Azure. To diversify its AI offerings and avoid dependence on a single provider, the company has now made Meta Platforms' open-source AI model, Llama 2, accessible through Azure AI Studio as a "model-as-a-service."

MaaS, akin to software-as-a-service (SaaS), allows customers to use Llama 2 on-demand via the web, significantly reducing the complexity of setup. Enterprises no longer need to worry about installing the model on their own cloud infrastructure—whether that be public, private, or hybrid. This is particularly advantageous, as deployments can be challenging even with expert assistance.

John Montgomery, Microsoft’s Corporate Vice President of AI Platform Program Management, explained in a blog post, “Azure AI has long allowed users to deploy models on their own infrastructure. However, not all customers want to manage this complexity. At Ignite, we introduced Models as a Service, which simplifies deployment by providing models as API endpoints, similar to the Azure OpenAI Service.”

In this public preview, Azure AI now offers a variety of Llama 2 models, including Llama-2-7b (Text Generation), Llama-2-7b-Chat (Chat Completion), Llama-2-13b (Text Generation), Llama-2-13b-Chat (Chat Completion), Llama-2-70b (Text Generation), and Llama-2-70b-Chat (Chat Completion).

Expanding AI Options for Customers

By introducing a diverse range of open-source Llama models, Microsoft enhances its Azure offerings, providing customers with cost-effective alternatives to OpenAI’s GPT-3.5 and GPT-4 models. Llama 2 has quickly become the preferred open-source choice for many users and enterprises in the generative AI space.

This strategic move also follows reports of CEO Satya Nadella expressing concerns over OpenAI's leadership changes and his desire to spread Microsoft’s AI investments more broadly.

Keeping OpenAI in the Mix

Despite the introduction of Llama 2, Microsoft continues to support OpenAI's advancements. Montgomery announced that OpenAI's latest model, GPT-4 Turbo with Vision—which can analyze and describe images—is now available to Azure customers as well. Companies like Instacart and WPP have started utilizing this model.

Additionally, Azure AI Studio provides tools for fine-tuning all available models. As the competition in the AI cloud space intensifies, the question remains: how many more models will Microsoft add to Azure AI Studio? Will Mistral or Deci be next to join the roster?

Earlier this week, Microsoft also released Phi-2, its own small language model (SML) for research purposes, although it is not yet available for commercial use. In contrast, Llama-2 and GPT-4 Turbo with Vision are ready for deployment in business applications.