Meta’s latest artificial intelligence models, Llama 4 Scout 17B and Llama 4 Maverick 17B, are now accessible through Amazon SageMaker JumpStart. These models will soon be available as fully managed, serverless options in Amazon Bedrock. The Llama 4 models are designed to process both images and text, offering advanced multimodal capabilities and larger context windows compared to their predecessors. This allows for more efficient and powerful applications, such as summarizing multiple documents or analyzing extensive user activity patterns.
The models’ innovative mixture of experts (MoE) architecture enables them to activate only the most relevant parts of the model for specific tasks, reducing computational resource usage while maintaining high performance. This makes advanced AI more accessible and cost-effective for businesses of all sizes. AWS continues to provide cutting-edge models from leading AI companies like Meta, ensuring customers have access to the latest technology with enterprise-grade tools and security.
Llama 4 Scout 17B boasts an industry-leading context window of up to 10 million tokens, nearly 80 times larger than previous versions. Llama 4 Maverick 17B excels in understanding images and text across 12 languages, making it ideal for sophisticated chat applications. Developers can now build applications that process vast amounts of information while supporting multiple languages and handling both text and images seamlessly.
AWS customers can expect continued expansion of models across sizes and modalities, empowering them to fully realize the potential of generative AI. Fully managed, serverless Llama 4 models in Amazon Bedrock will be available soon.
To start using the Llama 4 models, visit the Amazon SageMaker AI console.
— new from AboutAmazon.com