OpenAI opens access to GPT-4 API for everyone
OpenAI has made GPT-4, its advanced AI language model, available via a cloud-based API.
This is stated in the official OpenAI blog.
At the same time, the company provided third-party developers with access to the APIs of its other three AI models. In addition, OpenAI has announced its plans to gradually abandon a number of early neural networks.
The GPT-4 API allows developers to integrate the model into their programs.
The service is now available to all customers who have previously gained access to the API. The company will open access for new developers later this month.
In the future, OpenAI intends to increase the limits of API usage. The company’s long-term plans include the launch of a so-called fine-tuning capability that will allow developers to improve the accuracy of GPT-4 by training on user data sets.
Along with the GPT-4 API, access to the program interface of three other models was opened. The first is the GPT-3.5 Turbo, the predecessor to the GPT-4, which offers more limited features at a much lower cost. OpenAI also opened access to the APIs of its DALL-E and Whisper models. The latter two neural networks are optimized for image generation and speech transcription, respectively.
In connection with this event, OpenAI also announced plans to discontinue early language models available through the API. They will be disabled on January 4, 2024. Some of the models will be automatically updated to the new OpenAI products, while others will require developers to manually migrate.
The changes are partially related to the API update. Many of the models that will be phased out use an application programming interface that OpenAI calls the Completions API. It initially became available in 2020. For example, to connect to the Chat Completion API and run ChatGPT on a 39-year-old IBM computer, an enthusiast had to manually write the entire POST request in C. In March of this year, OpenAI introduced a new interface called Chat Completions API, which will be the main focus of further development.
“The Completions API was originally introduced in June 2020 to provide free-form text prompting for interacting with our language models. Since then, we’ve realized that we can often provide better results with a more structured tooltip interface,” OpenAI said in a blog post.
The new Chat Completions API brings several improvements. First of all, it provides protection against hinting attacks, which are attempts by hackers to change the output of the AI model using malicious instructions. This improvement should help developers improve the security of their OpenAI-based applications.
Earlier, OpenAI announced the release of new versions of GPT-3.5-turbo and GPT-4, introduced a new tool – function call, and announced a price reduction for developers.
