Microsoft launches Azure OpenAI service with ChatGPT coming soon

Microsoft is rolling out its Azure OpenAI service this week, allowing businesses to integrate tools like DALL-E into their own cloud apps. Microsoft has been testing this Azure service for just over a year, and it will soon include access to ChatGPT, the conversational AI that made headlines last year.
The Azure OpenAI service features a number of AI models made by OpenAI including GPT-3.5, Codex, and DALL-E, so businesses and developers can utilize these systems in their own apps and workloads. Microsoft essentially packages up GPT-3.5 with the scaling you’d expect from Azure and management and data handling additions. Developers could use Azure OpenAI to build apps that leverage AI for support tickets or for content matching to improve search results in online stores. Such models are already being used widely as tools to summarize documents and analyze text.



Would have been neat if they’d had a ChatGPT agent trained on the documentation ready to go

1 Like

Their pricing is a bit wonky. For fine-tuned models you pay both for the token usage and an hourly hosting fee. Those hourly costs could add up.

Also, is there such a thing as “GPT-3.5”? I hear this, but don’t see any such thing available via API, and have no idea how it differs from GPT-3 (aka how many parameters does “GPT-3.5” have?)

Finally, their embedding offerings look like the old ones, not the most recent ada-002 model.

1 Like

OK, here is what is considered “GPT-3.5” according to OpenAI. I had to do some digging:

Models referred to as “GPT 3.5”
GPT-3.5 series is a series of models that was trained on a blend of text and code from before Q4 2021. The following models are in the GPT-3.5 series:

code-davinci-002 is a base model, so good for pure code-completion tasks
text-davinci-002 is an InstructGPT model based on code-davinci-002
text-davinci-003 is an improvement on text-davinci-002




Is there any clear advantage of using Azure’s Open AI Service models over using OpenAI’s API on an AWS EC2? I mean any difference in quality of results, or can it be a case that users using Azure Open AI services will get newer models first rather than OpenAI API?

Actually I currently have my app deployed on an EC2 with AWS Activate credits. Do I need to move to Azure in near future or its just another way of getting the same results? Thanks.

1 Like

they don’t have the newer embeddings model :frowning:

1 Like