: A defining feature is the 10 million token context window available in some variants, allowing the model to "read" over 7,500 pages of text or process 20+ hours of video in a single prompt. Key Models in the Series
The Llama 4 series represents a major shift in open-source artificial intelligence, moving toward capabilities and Mixture-of-Experts (MoE) architectures. Laskamp4
: This is a larger model with 400 billion parameters and 128 experts. It rivals top proprietary systems like GPT-4 and Gemini in complex reasoning, coding, and image understanding. : A defining feature is the 10 million
: Unlike previous versions that relied on "bolted-on" vision components, Llama 4 was trained from the start with text, images, and video frames. It rivals top proprietary systems like GPT-4 and
: The models use a "mixture of experts," where only a subset of the total parameters (e.g., 17 billion active parameters in the Scout model) are activated for any given task. This significantly reduces computational costs and latency while maintaining high performance.