Llama 4 Scout

Hot

Meta's first open-weight natively multimodal MoE model with 16 experts and an industry-leading 10M token context window

Model Specifications

ArchitectureVISION

Parameters109B

Familyllama

VRAM (Q4)54.5GB

Mixture of ExpertsActive inference parameters: 17B.

Fits on a single H100 GPU with int4 quantization

#meta#moe#multimodal#long-context#trendingSource

Send this model's specs directly to your community.

Refined Llama 3 with superior following

Mobile-optimized small model

Ultra-light edge deployment

A complete breakdown of quantization levels and VRAM overhead for running local models.

Comparing NVIDIA and AMD options for the best speed-to-dollar ratio.

Understanding local AI formats and which one to pick for your specific hardware.