New AI Models on Amazon Bedrock: Llama 4, Ray2, and More

AWS has introduced Meta's latest foundation models, Llama 4 Scout and Llama 4 Maverick, on Amazon Bedrock and AWS SageMaker JumpStart, both featuring multimodal capabilities and a mixture-of-experts architecture. Llama 4 Scout is optimized for general-purpose tasks with 17 billion active parameters, while Llama 4 Maverick excels in reasoning and coding with the same number of parameters but spread across 128 experts, offering significant context length improvements over previous models. Additionally, AWS has integrated DeepSeek AI into its platforms, enhancing the training and deployment of machine learning models, with DeepSeek-R1 offering 671 billion parameters, emphasizing model safety with recommended guardrails. Luma AI's Ray2, a video-generating AI model, is available on Amazon Bedrock, enabling realistic video production via text prompts for various applications. Amazon Bedrock has also introduced new evaluation capabilities to streamline testing for generative AI, including RAG evaluation and LLM-as-a-judge functionalities. Furthermore, SSOJet offers secure SSO solutions tailored for B2B applications, featuring directory sync, SAML, OIDC, and magic link authentication to enhance enterprise authentication processes.