GPT-4o: text, voice and images in a single model
In one sentence OpenAI unveils GPT-4o (omni), a single model that natively handles text, audio, and images with 320 ms average voice response, GPT-4 Turbo-level text quality, and text+image capabilities opened to free users too.
OpenAI introduces GPT-4o ("omni"). The difference with prior models is that text, voice, and images no longer go through separate models: a single model handles everything.
Practical result, shown in the demo: you talk to ChatGPT the way you talk to a person, in real time, with interruptions and tone shifts, and you can show a photo and discuss it. Text and images with GPT-4o reach the free plan right away (with limits); the new voice mode is promised in alpha to Plus subscribers in the following weeks.
For developers or sysadmins, the API costs half as much as GPT-4 Turbo and is faster; native audio capabilities are opened to developers later.
Companies
OpenAI
Tools
GPT-4o, ChatGPT
Tags
Sources