OpenAI has introduced its new flagship generative model GPT-4o
GPT-4o – is a GPT-4 level model but faster and with new features:
• Image, audio and video recognition in real time (you can have a menu translated in a restaurant from a photo, or record a football match and ask the assistant to explain the rules of the game)
• Real-time voice translation (supports 50 languages)
• Voice assistant with natural speech: various emotions and intonations, whispering, singing
• Free for a limited number of requests
The number of queries for free use will vary depending on demand and load on the service at certain times. Once the limit is reached, users without a paid subscription will automatically be switched to GPT-3.5. With paid subscriptions, limits are increased.
The video provides examples of some new features (real-time voice translation and lullaby singing in a whisper), and even more demonstrations are posted in the thread X/Twitter OpenAI. Many of them are simply mind-blowing!
Not all features will be available right at the start: for example, the input of text and images in GPT-4o is implemented as of today, the voice and video will become available in the coming weeks.
Also, OpenAI has released a desktop application for macOS for more convenient access to ChatGPT. It is already available for Plus tariff subscribers. An application for Windows will appear later this year.
