According to the Financial Times, Meta
plans to introduce improved voice features in its next flagship large language model, Llama 4, which is expected to launch in a few weeks. The developers have paid special attention to the ability to interrupt and interrupt the model during a conversation, similar to OpenAI’s voice mode for ChatGPT and Google’s Gemini Live experience.
Image source: Pixabay
This week, Meta chief product officer Chris Cox said Llama 4 will be a “comprehensive” model capable of natively interpreting and outputting speech, as well as text and other types of data.
Image source: Meta
The success of Chinese AI lab DeepSeek’s open-source models, which have shown impressive results, has forced Llama’s developers to speed up their work significantly. Meta is even rumored to have set up war rooms to try to figure out how DeepSeek has managed to drastically reduce the cost of training, running, and deploying AI models.