Long story short:
- Desktop verstion.
- UI update.
- Launch of the new flagship model GPT-4o even for free users.
GPT-4o
Provides GPT-4 level intelligence, but it is much faster. Key point is ease of use. New model is shifting the paradigm of collaboration between ourselves and machines into the future. This interaction becomes easier and much more natural.
To understand voice, until now there were three models combined:
- transcription,
- intelligence,
- text to speech.
This brought a lot of latency to experience. With GPT-4o all of this happens natively.
New features
- Vision - upload screenshots, photos, images.
- Memory - sense of continuity across all conversations.
- Browse - search for real time information.
- Advanced data analysis - upload charts and other data sets.
- Improved quality and speed of 50 different languages, which is covering about 97% of world's population.
- Paid users will get 5x higher rate limits, compared to GPT-4 Turbo.
- New voice model features:
- You can now interrupt AI, while it's speaking.
- Real time response.
- The model picks up emotions.
- Model is capable of imitating different emotions, as well as sing and imitate different voices.
- Real time translation to multiple languages.
- New real time video model.
- ChatGPT desktop is capable of seeing live screen demonstration and answering on what's going on.
They will start to roll out this update today iteratively to all users.