ChatGPT developer OpenAI just announced it will enable access to its GPT-4 large language model (LLM) to all existing developers who have paid for the previous versions up until now.
“Today, all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability,” OpenAI noted in a statement on Thursday.
In addition, OpenAI is also making its GPT-3.5 Turbo, DALL-E, and Whisper application programming interfaces (APIs) broadly available, with older models of the Completions API retiring at the start of next year.
Commenting on its plans for future development, the company explained it will be labeling its previous GPT models as “legacy” in its developer documentation from July 6.
This means that the older GPT models will be considered outdated or no longer in active development or use, similar to an old version of the software that newer versions have superseded.
Its upcoming model and product improvements, meanwhile, will focus on Chat Completions API, OpenAI confirmed, adding that it doesn’t have plans to publicly release new models using the Completions API.
Starting January 4, 2024, older Completion models will no longer be available and will be replaced with new models available on the company's blog.
ChatGPT Now Significantly More Powerful With GPT-4
GPT-4, an upgrade from its predecessor, GPT-3.5, came with enhanced capabilities that include both text and image inputs, a marked progression from the previous text-only model. This model showcased "human-level" performance across various professional standards.
OpenAI is also currently testing an image-understanding feature using GPT-4 with Danish startup Be My Eyes that would allow users to generate or modify pictures using textual prompts. This could work similarly to the generative AI-powered content creation from Adobe, Firefly, which is already available in Photoshop.
GPT-4 was trained with publicly accessible data sourced from public web pages, along with data licensed by OpenAI. However, the company was recently accused of developing its products using stolen data.
Just last week, OpenAI faced a class action suit for allegedly misappropriating massive amounts of personal data from the internet. The lawsuit claimed that despite established protocols, OpenAI resorted to “theft” to systematically scrape 300 billion words from the internet without consent.