Llama 4 Scout is a 109 billion parameter Multimodal LLM (with 17 billion active parameters, using a mixture-of-experts architecture) designed for general-purpose use, including both text and image understanding. It excels in handling extremely long contexts—up to 10 million tokens—making it uniquely suited for tasks such as multi-document summarization, reasoning over vast codebases, and parsing extensive user activity or large document collections.

Some other noteworthy features of Llama 4 Scout include its native multimodality (jointly trained on text, images, and video), efficient inference thanks to its MoE design (16 experts with only a subset activated per token), and strong performance across reasoning, coding, multilingual tasks (supporting at least 12 languages), and visual question answering. Its architecture enables high throughput on single H100 GPUs without sacrificing quality or context length.

Metric	Value
Parameter Count	109 billion total / 17 billion active
Mixture of Experts	Yes
Active Parameter Count	17 billion
Context Length	10 million tokens
Multilingual	Yes
Quantized*	No

*Quantization is specific to the inference provider and the model may be offered with different quantization levels by other providers.