OpenAI has introduced the GPT-4o language model. It should be able to communicate with the user almost in real time via voice output. MacOS also gets a desktop app for ChatGPT.
ChatGPT developer OpenAI has introduced its new language model GPT-4o. It should offer capabilities at the level of the current GPT-4 model, but at the same time be faster and have better text, speech and image processing capabilities, as stated in the announcement.
The “o” in GPT-4o stands for “omni”, explains OpenAI – Latin for “all” or “whole”. The model is a step towards a more natural interaction between humans and computers. It accepts any combination of text, speech and images as input. It responds to audio inputs in an average of 320 milliseconds, with peaks of up to 232 milliseconds, it goes on to say.
The performance in English and with code is on a par with that of GPT-4, and even better in other languages, OpenAI promises. In the demo videos, the developers show how the tool communicates with a person in audio dialogue almost in real time. It can be interrupted while speaking without any problems.
GPT-4o is initially only for users of ChatGPT Plus and ChatGPT team available. Enterprise users should follow after that. The model will later be available to all users, but with restrictions. Plus users are said to have a message limit up to five times higher than Free users. With GPT-4o, free users will also have access to GPT-4, data analysis and diagram creation, image discussions and file uploads. GPT-4o will be rolled out in the coming weeks.
OpenAI also announced a ChatGPT desktop app for MacOS. This is intended to enable voice conversations directly from the computer. But the app is not yet available.
By the way: If you want to get better answers from ChatGPT, you should be nice to the chatbot – This is what research by the VMware NLP Lab shows.