Meta Launches Llama 4 AI Models to Counter DeepSeek Competition
Meta introduces Llama 4, a new family of multimodal AI models powered by an efficient mixture of experts architecture. The release includes two publicly available models and an updated Meta AI assistant, though strict licensing rules block EU users.
Meta releases Llama 4, a new flagship family of AI models that includes Scout, Maverick, and Behemoth. The tech giant trains these models on massive amounts of unlabeled text, image, and video data to provide broad visual understanding. This launch directly answers the rising competition from Chinese AI lab DeepSeek, whose efficient open models force Meta to accelerate its development timeline.
Scout and Maverick are now openly available on Llama.com and Hugging Face, while Behemoth remains in training. These models represent Meta's first use of a mixture of experts (MoE) architecture, which delegates processing tasks to smaller, specialized expert networks to boost computational efficiency. Maverick features 400 billion total parameters but only activates 17 billion at a time, making it highly effective for general assistant and chat tasks.
Meta updates its Meta AI assistant across WhatsApp, Messenger, and Instagram to run on Llama 4 in 40 countries, though multimodal features currently remain limited to English speakers in the U.S. The new license imposes strict geographic restrictions, explicitly prohibiting users and companies based in the EU from using the models due to regional data privacy laws. Additionally, companies with over 700 million monthly active users must apply for a special license that Meta grants at its sole discretion.