GPT-4 Turbo, OpenAI’s latest model, introduces 5 significant enhancements

During its recent developer conference, OpenAI made several significant announcements regarding its artificial intelligence tools, including ChatGPT. Notably, the company unveiled GPTs, a creator tool specifically designed for chatbots. Additionally, OpenAI introduced GPT-4 Turbo, a new model for ChatGPT. These two announcements stand out as the most noteworthy updates from the event.

OpenAI has previously introduced new models for ChatGPT, and this time is no exception. In the past, they upgraded ChatGPT from GPT-3.5 to GPT-4. If you’re wondering about the disparities that the upcoming GPT-4 Turbo version of the chatbot will bring, you’re not alone.

It is expected that the model will initially be made available to ChatGPT Plus subscribers before being released to the general public, following the pattern of previous launches.

OpenAI declined WIRED’s request for early access to the new ChatGPT model; however, we anticipate notable distinctions in GPT-4 Turbo.

New OpenAI Knowledge Cutoff

Eliminate the perpetual reminder from ChatGPT regarding its limited information cutoff date of September 2021 with ease. During a conference, Sam Altman, CEO of OpenAI, expressed his frustration with the fact that GPT-4’s knowledge of the world was limited to 2021, just like everyone else.

However, the new model now includes information up to April 2023, allowing for more current context when responding to prompts. Altman has vowed to prevent ChatGPT’s information from becoming outdated again. Nevertheless, the method of obtaining this information remains a contentious issue for authors and publishers who are dissatisfied with OpenAI’s use of their writing without their consent.

Input Longer Prompts: OpenAI

Feel free to extend and elaborate on your prompts without hesitation. According to Altman, GPT-4 Turbo has the capability to handle up to 128,000 tokens of context. It’s important to note that tokens do not directly correspond to the number of words in a prompt.

Altman drew a comparison between this new limit and approximately 300 pages worth of words in a book. This means that if you require the chatbot to analyze a lengthy document and generate a summary, you can now input a larger amount of information in a single prompt using GPT-4 Turbo.

Enhanced Compliance with Directions

It would be desirable if ChatGPT possessed enhanced capabilities in comprehending the intricate specifics of the user’s prompt. OpenAI has stated that the forthcoming model will exhibit improved attentiveness.

OpenAI’s blog post states, “GPT-4 Turbo outperforms our prior models in tasks that necessitate meticulous adherence to instructions, such as generating precise formats (e.g., ‘always respond in XML’).” This advancement could prove especially advantageous for individuals who rely on the chatbot’s aid in coding.

Prices for developers at a more affordable rate

The cost of utilizing the application programming interface (API) provided by OpenAI may not be a primary concern for the majority of ChatGPT users. However, it is important to note that developers may find it relatively expensive.

Altman explained that the revised pricing structure entails a charge of one cent for every thousand prompt tokens and three cents for every thousand completion tokens. In simpler terms, this implies that developers may incur lower expenses when utilizing GPT-4 Turbo to input data and obtain responses.

Chat with Multiple Tools Integrated

ChatGPT Plus subscribers may have previously utilized the GPT-4 dropdown menu to choose their preferred chatbot tools. This menu allowed users to select options such as the Dall-E 3 beta for AI-generated images or the Browse with Bing version for internet links. However, this dropdown menu will soon be discontinued.

According to Altman, “We have taken your feedback into consideration and have decided to eliminate the model picker as it was deemed extremely bothersome.” The updated chatbot with GPT-4 Turbo will now automatically select the appropriate tools based on your request. For instance, if you ask for an image, the chatbot will use Dall-E 3 to fulfill your prompt.