Thursday, OpenAI he announced that all customers who pay for the API now have access to the GPT-4 API. It also introduced changes to chat-based models, announced the transition from the Completions API to the Chat Completions API, and outlined plans to retire older models.
Often considered to be the most powerful API tool, the GPT-4 API was implemented for the first time March but it has been a closed trial until now. As an API, developers can use special features to integrate OpenAI’s main language (LLM) into their applications such as abstraction, document support, analysis, and design. The model runs remotely on OpenAI servers and provides output to other applications over the Internet.
OpenAI says the GPT-4 API is 8K content it is accessible to existing developers with a good payment history, with plans to open the opportunity to new developers at the end of July. And in a move away from the old GPT-3 versions, OpenAI has also decided to abandon the “Completions API” versions in favor of new ones. Chat Completions API examples. Since March to initiateOpenAI says its version of the Chat Completions API now accounts for 97 percent of OpenAI’s API GPT usage.
This retirement plan includes retiring samples that are part of the Completions API in six months. As of January 4, 2024, these old models, which are fast-ending rather than chat-like, will be instead and newer versions, and developers using some of these models will need to upgrade their integrations manually. For now, the old models will still be available but will be listed as “legacy”. Here is a list of some of the types that will be removed:
OpenAI will provide input for these old models. For example, OpenAI recommends that users a text-davinci-003 change the version to the gpt-3.5-turbo-instruct version instead. Similarly, the versions ada-002, babbage-002, curie-002, and davinci-002 will replace the previous versions of each. Software using some of the GPT-3 versions (such as ada, babbage, curie, davinci) will “automatically be upgraded to the new versions listed above on January 4, 2024,” according to OpenAI.
OpenAI also announced that “based on the stability and readiness of these models for use in production,” it is also developing APIs for Whisper, GIVE HIM, and GPT-3.5 Turbo is “frequently available.” And the company hopes to continue to improve these models throughout the year.
Developers can find more information in OpenAI’s blog post about the announcement.