OpenAI, under the leadership of Sam Altman, unleashed the newest version of its GPT-4 model, 'GPT-4o', designed to generate diverse outputs—text, audio, and images—on Monday.
OpenAI has rolled out the GPT-4o text and image features to all ChatGPT users, regardless of subscription level. During a live-stream event, it was revealed, "We are launching GPT-4o in the free tier and extending it to Plus users with up to 5 times higher message limits."
The company also announced that it will soon be launching an Alpha version of Voice Mode with GPT-4o in ChatGPT. Sam Altman, while excited about the new model, said, "Our new baby, GPT-4o, is the pinnacle of our progress. It's smart, fast, and natively multimodal."
Altman added it is broad-based: "It's available to all ChatGPT users—even users on the free plan! Traditionally, access to GPT-4 class models has only been available to subscribers. But democratizing access to powerful tools for AI is part of our mission."
GPT-4o is very responsive; it processes audio inputs in 232 milliseconds and responds in 320 milliseconds on average—the pace at which human beings converse. Its remarkable ability to understand visual and audio cues is unmatched compared to its predecessors.
GPT-4o's enhanced audio and video capabilities will be supported and rolled out with a select group of trusted partners using the API over the coming weeks.
GPT-4o is unique in that it has used one unified approach across text, vision, and audio. OpenAI trained a single model end-to-end, ensuring that all inputs and outputs go through the same neural network. "As GPT-4o represents our first model that integrates these modalities, we are just getting started exploring what it can and can't do," OpenAI said.
The company also launched a Mac desktop application for ChatGPT alongside the event. Besides this, OpenAI released the GPT Store, which is now freely available to users and allows them to build and distribute their custom chatbots, known as GPTs.
Read also | Elon Musk Criticizes OpenAI's Latest Demo as Cringe-Worthy
