ChatGPT developer Open AI has made the GPT-4 API generally available for all existing OpenAI API developers with a track record of successful payments.
This means that developers can harness the power of GPT-4 to augment their applications to have advanced AI features – leading to a new wave of innovative applications in the market.
The company plans to extend this access to new developers by the end of this month, with gradual increments to availability limits dependent on computing capacity.
"Millions of developers have requested access to the GPT-4 API since its launch in March, and the spectrum of innovative products leveraging GPT-4 is expanding daily," OpenAI stated in a blog post.
With this milestone, Open AI envisions a future where chat-based models can accommodate any use case.
Significant upgrade
GPT-4 offers improvements over its predecessor, GPT-3.5, as it can generate text (including code) and accept both image and text inputs.
This marks a significant upgrade since GPT-3.5 only accommodated text inputs. OpenAI revealed that GPT-4 has demonstrated "human-level" performance on diverse professional and academic benchmarks.
The model was trained using publicly accessible data, comprising public web pages and data that OpenAI licensed.
OpenAI has yet to roll out the image-understanding feature to all its customers.
Currently, it is being tested with a single partner, Be My Eyes. However, the company has not confirmed when it will extend this feature to its wider customer base.
GPT-4 API is now available to all paying OpenAI API customers. GPT-3.5 Turbo, DALL·E, and Whisper APIs are also now generally available, and we’re announcing a deprecation plan for some of our older models, which will retire beginning of 2024: https://t.co/H11w4IalNZ— OpenAI (@OpenAI) July 6, 2023
"Hallucinates" facts
Despite its advanced capabilities, GPT-4 has limitations. The AI sometimes "hallucinates" facts, makes reasoning errors confidently, and doesn't learn from its mistakes.
For instance, it fails to address complex problems, such as introducing security vulnerabilities into the code it generates.
In upcoming developments, OpenAI plans to allow developers to fine-tune GPT-4 and GPT-3.5 Turbo, another of its less capable yet recent text-generating models, with their own data.
This adjustment capability is expected to be available later this year.
Seamless transition
In a bid to ensure a seamless transition, OpenAI pledged to provide ample support to users who had previously fine-tuned models.
"In the coming weeks, we will reach out to developers who have recently used these older models and will provide more information once the new completion models are ready for early testing," OpenAI wrote.
Developers will need to manually upgrade their integrations by the specified date to continue using fine-tuned old models and will need to fine-tune replacements using the new base GPT-3 models.