Microsoft drastically expands Azure AI Studio to include Llama 2 Model-as-a-Service, GPT-4 Turbo with Vision

Microsoft, one of the biggest winners of the generative AI boom so far thanks to its early backing of OpenAI and integration of the latter startup’s tech into Bing, Azure, and various other services, has been trying to avoid putting all of its AI eggs in one basket.

Today, the company announced that it is making Meta Platforms’ rival open-sourced AI model Llama 2 available in Azure AI Studio, its cloud platform AI clearinghouse, in the form of a “model-as-a-service.”

Understanding AI models-as-a-service

Similar to the now-old acronym of SaaS (software-as-a-service), MaaS offers customers the ability to use AI models such as Llama 2 on-demand over the web when needed with minimal setup. They won’t have to go through all the trouble of installing it on their own cloud server space, private cloud, hybrid cloud, or other locations. Even with Microsoft experts helping and trained IT personnel, this can be a non-trivial task for enterprises.

As John Montgomery, Microsoft Corporate Vice President of Program Management for its AI Platform efforts explained in a blog post:

“In Azure AI, you have been able to deploy models onto your own infrastructure for a long time – simply go into the model catalog, select the model to deploy and a VM to deploy it on and you’re off to the races. But not every customer wants to think about operating infrastructure, which is why at Ignite we introduced Models as a Service, which operates models as API endpoints that you simply call, much the way you might call the Azure OpenAI Service.

Today, we’re making Meta’s Llama 2 available in Models as a Service through Azure AI in public preview, enabling 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).”

Hedging AI bets while giving customers more choices

Offering this wide range of different open-source Llama models seems like a smart move on Microsoft’s part, expanding the choices of AI for Azure cloud storage and service customers and giving them a far lower-cost option (the models themselves are free, though various implementations may not be) to Microsoft’s buddy OpenAI’s GPT-3.5 and 4 models.

After all, Llama 2 arguably won the year when it comes to generative AI since it is the preferred open-source option for many users and enterprises.

It follows reports of Microsoft CEO Satya Nadella, who was said to be surprised and furious about the prior OpenAI board’s abrupt decision to terminate CEO Sam Altman back in November before ultimately re-hiring him, stating that he wants to spread Microsoft’s AI bets more widely.

OpenAI still well-represented

But that doesn’t mean that OpenAI is being left out in the cold with today’s announcements. On the contrary: Montgomery also shared the news that OpenAI’s latest publicly accessible AI model, GPT-4 Turbo with Vision — which gives the AI the capabilities of analyzing and describing photos and visual material — is also now available to customers of Azure in the Azure AI Studio.

Already, customers such as Instacart and WPP are using this model, according to Montgomery.

And, the Azure AI Studio includes tools for fine-tuning all the models offered.

At this point, as the AI cloud wars continue to rage on, the only real question is: how many models will Microsoft be able to add to Azure AI Studio? Is Mistral or Deci next?

The news today also follows Microsoft’s release earlier this week of Phi-2, its own small language model (SML), though this is only for research purposes and can’t be used for commercial cases. However, Llama-2 and GPT-4 Turbo with Vision can.