SageMaker JumpStart now offers optimized deployments for foundation models

Amazon Web Services has launched optimized deployments for SageMaker JumpStart, a new feature that enables customers to deploy foundation models with pre-configured settings tailored to specific use cases and performance requirements. The service offers task-aware configurations that can optimize for cost, throughput, or latency depending on workload needs such as content generation, summarization, or question-and-answer applications. The launch includes support for more than 30 popular models from major AI providers including Meta, Microsoft, Mistral AI, Qwen, Google, and TII, with pre-deployment visibility into key performance metrics like P50 latency, time-to-first token, and throughput. Customers can select from use case-specific configurations for scenarios like generative writing or chat-style interactions, then choose optimization targets including cost-optimized, throughput-optimized, latency-optimized, or balanced performance options. Supported models include Meta's Llama 3.1 and 3.2 variants, Microsoft Phi-3, Mistral AI models including the new Mistral-Small-24B-Instruct-2501, Qwen 2 and 3 series with multimodal Qwen2-VL capabilities, Google Gemma, and TII Falcon3. The models deploy to SageMaker AI Managed Inference endpoints or SageMaker HyperPod clusters with enterprise-grade security through VPC deployment capabilities, and the feature is available across all AWS regions where SageMaker JumpStart currently operates.