Meta is preparing to release a new version of its generative AI model, Llama 4, which reportedly will focus on voice features. The new model is expected to arrive within a few weeks. The main feature of Llama 4 will be the ability for users to interrupt the model while it is speaking, similar to the Voice Mode feature from OpenAI for ChatGPT and Gemini Live from Google.
At the Morgan Stanley conference, Meta’s Chief Product Officer, Chris Cox, noted that Llama 4 will be an “omni” model, capable not only of processing text but also interpreting and generating speech, as well as other types of data. This opens up new opportunities for users, who will be able to interact with the model in a more natural way.
The development of Llama 4 has accelerated amid the success of open models from the Chinese AI lab DeepSeek, which demonstrate performance on par with or even better than Meta’s models. In response, Meta has made efforts to study the methods that allowed DeepSeek to reduce the cost of launching and deploying models.
The new Llama 4 model from Meta promises to significantly expand user capabilities thanks to improved voice features, which could become an important step in the development of AI technologies.
