Meta Releases Llama 4 Scout and Maverick with 10M Token Context Windows

Meta released Llama 4 Scout (17B active / 109B total MoE parameters) and Llama 4 Maverick (17B active / 400B total MoE parameters), both featuring 10 million token context windows and open-weight Mixture-of-Experts architecture. Scout targets broad use while Maverick focuses on multilingual and coding workloads. Both are open-source on Llama.com and Hugging Face, with Meta AI updated across WhatsApp, Messenger, and Instagram Direct in 40 countries, strengthening Meta's position in the open-source AI ecosystem.