Meta Launches Llama 4 AI Models, Introducing Scout and Maverick

The new models, leveraging multimodal capabilities and mixture-of-experts architecture, are now available, while the larger Behemoth model remains in training.

Meta has released Llama 4 Scout and Maverick, its latest AI models, which are now integrated into WhatsApp, Messenger, and Instagram Direct.
The models use a mixture-of-experts (MoE) architecture, improving computational efficiency by activating only relevant parts of the model for specific tasks.
Scout is optimized for long-context tasks with a 10-million-token context window, while Maverick excels in general assistant use cases such as creative writing and coding.
Meta claims Maverick outperforms OpenAI's GPT-4o and Google's Gemini 2.0 on benchmarks for reasoning, coding, and image analysis.
Llama 4 Behemoth, still in training, is expected to feature 288 billion active parameters and surpass competitors on STEM-related benchmarks.