OpenAI has finally upgraded ChatGPT to run on its new GPT-4-Turbo model. This brings with it improved reasoning capabilities, a more conversational tone and overal better performance.
In recent weeks the AI chatbot had become somewhat lazy and more direct in its responses but OpenAI says the new upgrade will solve those issues.
The latest version of GPT-4-Turbo will only be available to ChatGPT Plus subscribers. Those using the free version of the chatbot or using it without an account still only have GPT-3.5.
OpenAI says its updated model has “improved capabilities in writing, math, logical reasoning, and coding,” as well as more up to date information with a December 2023 knowledge cut off.
What is new in GPT-4-Turbo?
Our new GPT-4 Turbo is now available to paid ChatGPT users. We’ve improved capabilities in writing, math, logical reasoning, and coding. Source: https://t.co/fjoXDCOnPr pic.twitter.com/I4fg4aDq1TApril 12, 2024
GPT-4-Turbo is the latest version of OpenAI’s incredibly powerful GPT-4 family of large language models. It is also the first to include AI vision capabilities, allowing the model to analyze the contents of a video or live feed for the first time.
It was released to developers earlier this week and has now found its way into OpenAI’s flagship consumer and enterprise product ChatGPT.
As an indication of how much better Turbo is compared to previous releases, it has put OpenAI back at the top of the LMSys AI leaderboard. This is a chatbot arena that lets users vote on the best output in a blind test of two models.
GPT-4 had held the top spot for a long time until Anthropic released Claude 3, getting knocked down to second place. It is now back at the top with users highlighting its natural responses.
What is new in ChatGPT?
Bringing GPT-4-Turbo into ChatGPT so soon after its developer release is a strong signal from OpenAI of its confidence in the chatbot. It is also the worst the AI model will ever be.
William Fedus, a researcher at OpenAI said on X that there were ”better versions to come”. This mirrors comments from CEO Sam Altman that there are some impressive and powerful AI models launching this year — potentially including Sora and GPT-5.
When writing with ChatGPT, responses will be more direct, less verbose and use more conversational language
OpenAI
OpenAI put its GPT-4-Turbo model through a series of evaluations and benchmarks before release finding increased performance on the majority of counts and significantly increased performance on GPQA — a graduate-level question and answer benchmark set by Google.
“When writing with ChatGPT, responses will be more direct, less verbose and use more conversational language,” OpenAI said in a statement.
They plan to continue to invest in making the models better and in turn improving how well ChatGPT performs. It should be noted GPT-4-Turbo is already available in Microsoft Copilot for both free and paid users — although it is an earlier version.
