OpenAI has revealed its latest language model, GPT-4 is now generally available to paying API customers. In an official blog post, the company outlined the general availability of GPT-4, along with its plans to retire several APIs by January next year.
Existing API developers who have made payments to OpenAI can now access GPT-4 with an 8K context, equivalent to approximately 13 pages of text. OpenAI also plans to raise rate limits and open access to GPT-4 to new developers by the end of July, based on compute availability.
OpenAI further emphasized the stability and preparedness of the GPT-3.5 Turbo, DALL-E, and Whisper models for production-scale use in the blog post, announcing their general availability. It indicated that work is being done to securely enable fine-tuning for GPT-4 and GPT-3.5 Turbo, with a release later this year predicted.
Furthermore, OpenAI indicated that GPT-3 and its associated variants utilizing the Completions API would be phased out beginning January 4, 2024, in order to free up processing space. While the API will remain available, it will be labelled as “legacy” and will eventually be replaced by a new model. According to OpenAI, it aims to focus on the Chat Completions API for future models and product developments, with no plans to offer new models utilizing the Completions API.
The sources for this piece include an article in ZDNet.
