OpenAI makes the API platform for its GPT models more secure and cost-effective, making it more attractive to enterprise users.

Competition in the AI ​​world isn’t getting any easier for OpenAI, and so the company is stepping up. A blog explains some new features for enterprise customers who use APIs to connect the GPT models to their business environment. OpenAI knows which sensitive areas need to be touched: security, control and cost management.

Private links

To increase the security of APIs, OpenAI introduces Private links. This feature allows users to create a direct connection between the models and their Azure cloud to reduce the burden on the public internet. OpenAI also adds MFA to manage access to models.

The ProjectsThe feature is intended to give companies a better overview of how models are used within the organization. In the dashboard, you assign roles and APIs to specific projects and determine which models you want to make available for each project. Project owners also have the ability to create API keys for service accounts, allowing access to projects without being tied to a single user.

Discounts

Using APIs is not free. Two new features are intended to ensure that companies’ AI accounts are not bursting at the seams. Businesses that utilize a constant level of tokens per minute on GPT-4 and GPT-4 Turbo may qualify for a “provisioned throughput” discount of ten to fifty percent, depending on the size of the commitment.

This seems like a pretty complex calculation, but OpenAI offers discounts in other ways. Customers can now also use it Batch API to run non-urgent AI workloads asynchronously. These queries provide a result within 24 hours at a reduced rate of fifty percent.

Finally, OpenAI is also looking at the Assistants API, a tool that helps companies search for custom models. The API tool will include improved search capabilities, file management, real-time streaming, and control over token usage.