"Laskamp4" appears to be a specific term likely referring to the lama 4 S cout model or related developments within the Llama 4 series released by Meta AI in April 2025 . The Llama 4 Series Overview
The Llama 4 series represents a major shift in open-source artificial intelligence, moving toward capabilities and Mixture-of-Experts (MoE) architectures. Laskamp4
: Designed for efficiency, this model has 17 billion active parameters. It fits on a single H100 GPU. It is optimized for high-speed performance (up to 460+ tokens per second) and long-document reasoning. "Laskamp4" appears to be a specific term likely
: 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. It fits on a single H100 GPU