OpenAI Presented an Improved GPT-4 Turbo Model

Author at ApiX-Drive

Reading time: ~2 min

Recently, representatives of OpenAI held a conference for developers, during which they presented GPT-4 Turbo, their latest AI product. As the name suggests, this is an improved version of the well-known and popular language model. OpenAI noted that the new product is a more efficient and economical alternative to GPT-4.

Functionality and differences of the GPT-4 Turbo:

Increased prompt size. Now you can write a hint to the chatbot up to 128,000 tokens long. Translating these units of measurement into a more understandable format, we can say that the longest clue would take up 365 pages of the book. For greater clarity, imagine the novel “Harry Potter and the Philosopher’s Stone” by J. K. Rowling, which has approximately the same length.
The model can write a response to the user in JSON text format.
Added the ability to upload your documents (for example, in PDF format).
The language model database has been “pumped up” until April 2023. This means that the neural network is familiar with all the events that occurred before this time. Accordingly, the user can count on more informative and factually accurate answers from the chatbot.
Significant savings when working with APIs for developers. In the new version of the language model, the cost of a prompt is 3 times lower than in the previous one, and the cost of a generation is 2 times lower.
Constructor of personalized chatbots. GPTs are personal AI assistants designed for specific tasks. Any user will be able to create such an assistant directly in the ChatGPT dialog box without any knowledge or coding skills.
Own application store. GPTs can be shared publicly. Those who develop plugins or chatbots will be able to make money here by receiving a portion of the income from subscribing to them.
Text-to-speech functionality using the speech synthesis API. The new version of the language model provides 6 preset naturalistic human voices.

Currently, there are 2 versions of GPT-4 Turbo: one analyzes text information, and the other is endowed with the ability to understand the context of not only texts but also images. OpenAI promises to provide general access to both in the coming weeks.