OpenAI has been creating a lot of buzz as of late with it’s impressive models and public access to these models. Lots of businesses are trying to leverage the power of generative AI to improve their solutions. OpenAI and Microsoft have teamed up to create Azure OpenAI, which allows Azure customers to quickly, easily, and securely implement AI capabilities into their application. The real question is, which option is right for you, using the OpenAI API or using Azure OpenAI? Here we will cover the differences between OpenAI and Azure OpenAI to help choose the right option for you.
Firstly, it’s important to note that both of these products are rapidly evolving and the information in this article may become outdated over time. It’s important to review the most up to date pricing for each product that can be found here:
Both of these products will provide API access to the OpenAI models allowing you to prompt and get responses as expected. Both services use a usage based pricing model that way you only pay for what you use. OpenAI and Azure OpenAI’s pricing models are calculated by how many tokens are used and both have similar pricing for interacting with the models. Both options also allow you to fine tune models with your own training data and set token usage limits.
Currently OpenAI has the following features that Azure OpenAI doesn’t:
- Access to the Whisper (audio) model.
- The option to generate lower resolution images using DALL-E that have lower price points (Azure Open AI only generates images in the highest resolution).
- 1024×1024, 512×512, 256×256.
- Provides Python and Node.js libraries to access the API.
- Libraries for other languages are available but are community maintained.
- OpenAI library information can be found here.
Currently Azure OpenAI offers the following features that OpenAI doesn’t:
- Embedded Azure platform and model security.
- Access to Codex models that are used for code generation.
- Models can be fine tuned and are hosted by Azure meaning your private training data and models stay private.
- Requires an Azure subscription to access.
- Provides a Python and C# SDK to access the service from your application.
Which product you decide to go with really comes down to a few things simply because the pricing is so comparable.
It’s recommended to use Azure OpenAI if you are already using the Microsoft stack and have an Azure subscription. In the case of Great Ideaz and our trellispark platform, we’re adding Azure OpenAI to use in trellispark for the following reasons:
- We already use Azure services and have a subscription.
- Our solution is built on the Microsoft stack.
- The C# SDK provides a very easy setup and integration into our system.
- The security provided by Azure to keep our data safe and moderation tools to prevent model abuse is important to us.
- Allows us to securely tune models with our own training data.
- The ability to take advantage of the Codex models used for code generation.
If you’re not on the Microsoft stack and don’t have an existing Azure subscription, the OpenAI API available directly through OpenAI would be recommended. Especially if you’re just an enthusiast that wants to try out generative AI in a small project, the OpenAI API is a great option. If you’re not using python or node.js, you can still use one of the community maintained libraries. Azure definitely adds some additional features, structure, and security, but that’s not always required to have depending on your business needs. At Great Ideaz, we intend to add integrations with multiple generative AI solutions into our trellispark platform, so our customers can choose from a range of AI products to enrich their user experiences.