OpenAI announced new features aimed at enterprise users on Tuesday, at the wake of Meta’s recent LLama 3 announcement.

The features include Private Link, which allows customers to establish direct communication between Azure and OpenAI while reducing exposure to the open internet.

Projects is another new feature, and provides organizations control over individual projects in OpenAI. This includes the ability to scope roles and API keys to specific projects, restrict or allow access to certain models, and set usage and rate-based limits, says OpenAI. Project owners can also create service account API keys, which grant access to projects without being tied to individual users.

OpenAI has introduced updates to the Assistants API, including retrieval with ‘file_search’, which can ingest up to 10,000 files per assistant. The API now supports streaming for real-time responses and includes new ‘vector_store’ objects. Users can control the maximum number of tokens used per run and select specific tools for each run. The API also supports fine-tuned GPT-3.5 Turbo models.

There are new pricing options as well, with discounted usage on committed throughput and reduced costs on asynchronous workloads. Customers with consistent usage on GPT-4 or GPT-4 Turbo can request access to provisioned throughput to receive discounts based on the size of their commitment. The new Batch API allows customers to run non-urgent workloads asynchronously at a 50% discount, with results returned within 24 hours.

OpenAI enterprise customers include Klarna, Morgan Stanley, Oscar, Salesforce, and Wix.

[Image courtesy: OpenAI]