Llama 4 Maverick

Llama 4 Maverick is a large language model Meta released on 5 April 2025 as part of the Llama 4 family. It uses a mixture-of-experts architecture with 128 experts, roughly 400 billion parameters in total and 17 billion active per token, takes text and image input natively, and is documented with a context window of up to one million tokens. It was codistilled from Llama 4 Behemoth and is distributed as open weights under the Llama 4 Community License, which carries use restrictions, including attribution requirements and a separate permission requirement from Meta above 700 million monthly active users. Within the family, Maverick sits above Llama 4 Scout: both activate 17 billion parameters per token, but Maverick spreads them across 128 experts instead of 16, which raises total memory requirements while keeping compute per token comparable. Meta positioned it as a general-purpose assistant and chat model, available through cloud providers and self-hostable on a single NVIDIA H100 DGX host using the FP8 weights. Typical uses are assistant and chat products, document and image understanding, coding support, and long-context retrieval workloads. Maverick is a released production model rather than a research preview; within Llama 4, the model that stayed internal is the unreleased Behemoth.

Related terms

• Code Llama
• LangChain Llama 3
• Llama 3
• Llama 4 Behemoth
• Llama 4 Scout

Related services: Agentic AI consulting.

← Back to the AI Glossary