Tue. May 21st, 2024



ChatGPT is powered by a few different AI models, with GPT-4 Turbo for paying subscribers and GPT-3.5 for everyone else. OpenAI has now revealed a new model that will soon be available for everyone: GPT-4o.



GPT-4o (the “o” is apparently for “omni”) is a new AI model that handles text, audio, or images as both input and output. That’s a big change from OpenAI’s previous models, which were more specialized and had to be combined for different workloads. Google Gemini is designed to work in a similar way, so it seems like OpenAI is catching up to the competition.

In a blog post, OpenAI said GPT-4o “can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models.”


The upgraded sound and voice capabilities mean GPT-4o can behave more like a virtual assistant, which is something OpenAI and Google have tried with mixed success in the past. OpenAI plans to add a Voice Mode “in the coming weeks” powered by GPT-4o, which will be available first to ChatGPT Plus subscribers.

GPT-4o isn’t a full replacement for GPT-4 and GPT-3.5, at least not yet. It might behave differently in some scenarios, since it is trained differently, but OpenAI is promising similar performance to GPT-4 Turbo in text, reasoning, and coding intelligence. Importantly, it will be available on free ChatGPT accounts, which have been limited to the less-capable GPT-3.5 model until now.


The new GPT-4o model is rolling out now in ChatGPT, for both free users and ChatGPT Plus subscribers. It will have a message limit, but the limit will be “5x higher” for Plus users. ChatGPT Plus subscribers will also get access to the Voice Mode “in the coming weeks,” and the new model is now available as an API for developers to try out.

Source: OpenAI (1, 2)



Source link

By John P.

Leave a Reply

Your email address will not be published. Required fields are marked *