Meta released Llama 4 as open-source software, continuing its strategy of making powerful language models available to the broader developer community. The release includes three model sizes (8B, 70B, and 405B parameters) with a common 2M token context window.
The standout feature is what Meta calls "native context engineering" — a set of built-in primitives for structured context management that previously required external frameworks. These primitives handle retrieval, memory management, and tool context within the model architecture itself.
The licensing remains more permissive than most "open-source" LLM licenses, with specific commercial-use clauses that make Llama 4 genuinely usable for enterprise deployment without navigating complex legal restrictions.
What this means: The gap between proprietary and open-source LLM capabilities continues to narrow. Llama 4's 2M token context window and native context engineering put it within striking distance of GPT-5.1 and Claude Enterprise for many enterprise use cases — at zero licensing cost.
