According to the Financial Times, Meta is set to launch its next flagship large language model, Llama 4, in the coming weeks. This version will introduce enhanced speech capabilities, with a focus on enabling users to interrupt and resume conversations naturally. These improvements aim to bring Llama 4 closer to OpenAI’s ChatGPT speech mode and Google’s Gemini Live experience.
Meta’s chief product officer, Chris Cox, recently stated that Llama 4 will be a “comprehensive” model, capable of interpreting and inferring speech, text, and other types of data natively. This suggests a significant leap in multimodal AI capabilities, making interactions more fluid and natural.
Competition with DeepSeek and Cost Reduction
The rise of open-source models from the Chinese AI lab DeepSeek has increased pressure on Meta to accelerate its development efforts, adds NIXsolutions. DeepSeek’s AI models have delivered impressive results while drastically reducing the cost of training, operation, and deployment. In response, Meta has reportedly set up dedicated teams—sometimes referred to as “war rooms”—to analyze DeepSeek’s strategies and optimize its own processes.
Llama 4’s upcoming release marks an important step in Meta’s AI ambitions, aiming to enhance real-time speech interaction while staying competitive in the evolving AI landscape. We’ll keep you updated on further developments.