OpenAI Lowers Prices And Presents New Features At GPT-3.5 Turbo And GPT-4

JAKARTA - In the midst of increasing competition in the generative Artificial Intelligence (AI) room, OpenAI released new features and price drops for its text-producing models GPT-3.5 Turbo and GPT-4.

It is known, the GPT-4 model is the most advanced version today and the latest, it can interact better with third parties, and lower prices on other ChatGPT functions.

Another new feature added to GPT-4 and GPT-3.5 Turbo is functional summons. With this, users can describe the programming function to bots and intuitively create codes to realize such functions.

For example, when users ask what the weather is like in Jakarta, and it will go through the whole process to give them an answer.

These models have been adjusted to detect when a function needs to be called and to respond with the JavaScript Object Notation (JSON) which follows the signature of the function," OpenAI said in a blog post, quoted Friday, June 16.

"The summoning of functions allows developers to regain structured data from the model more reliably," he added.

OpenAI also extended the context of GPT-4 for a better understanding of text. The company will call in more of the waiting list to test its features.

"With this update, we will invite more people from the waiting list to try GPT-4 over the coming weeks, with the intention of deleting the entire waiting list with this model," OpenAI said.

While GPT-3.5 Turbo will have the same functional summons as GPT-4 as well as more reliable steering capabilities through system messages, two features that are beneficial for developers to guide model responses more effectively.

This new version of GPT-3.5 Turbo costs twice the previous version, but the good news is that the old version will get a 25 percent discount.

"All of these models come with the same privacy and data security guarantees that we introduced on March 1, customers have all the outputs generated from their requests and their API data won't be used for training," OpenAI concluded.