AWS releases
Amazon Web Services releases and Terraform AWS provider. New features, breaking changes, security advisories and deprecations - each summarised in plain English and updated continuously.
Tracking 284 AWS releases · Updated
- AWS What's New aiawsgaengineer ·
Amazon OpenSearch Serverless Next Generation GA
AWS announced the general availability of the next generation of Amazon OpenSearch Serverless, featuring 20x faster auto-scaling and resource provisioning in seconds. This fully managed search and vector engine offers scale-to-zero pricing, potentially saving customers up to 60%, and decouples compute and storage for independent scaling. It also simplifies network connectivity with new endpoints and offers native integrations with AI development platforms, making it available today in all commercial AWS regions.
feature announcement - AWS What's New aiawsengineer ·
Amazon Connect Assistant integrated into UI builder
Amazon Connect's AI assistant is now available within the UI builder, allowing contact center managers to create and modify Views using natural language prompts. This feature can reduce the time needed to build Views by up to 70%, enabling managers to configure layouts, components, and styling conversationally. The assistant also provides recommendations and troubleshooting, speeding up the development process.
feature - AWS What's New aiawsengineer ·
P5.48xl instances with NVIDIA H100 GPUs now available on AWS SageMaker in Tokyo
Amazon SageMaker notebook instances now support P5.48xl instances with NVIDIA H100 GPUs in the Asia Pacific (Tokyo) region. These instances offer up to 4x faster deep learning and HPC performance compared to previous generations, reducing ML model training costs by up to 40%. They are suitable for training and deploying large language and generative AI models, with general availability starting now.
feature announcement - AWS What's New aiawsgaengineer ·
AWS SageMaker adds NVIDIA Blackwell P6-B200 instances for AI training
Amazon SageMaker notebook instances now offer general availability of EC2 P6-B200 instances, featuring 8 NVIDIA Blackwell GPUs and Intel's Emerald Rapids processors. These instances aim to accelerate AI training, offering up to twice the performance of P5en instances for tasks like fine-tuning large foundation models. This enhancement is particularly beneficial for developers working with LLMs and generative AI applications, providing more power for interactive development directly within JupyterLab or CodeEditor environments.
feature announcement - AWS What's New aiawsengineer ·
P4de instances available on SageMaker Notebook Instances in Tokyo
Amazon EC2 P4de instances, featuring 8 NVIDIA A100 GPUs with 80GB HBM2e memory each, are now generally available on SageMaker notebook instances in the Asia Pacific (Tokyo) region. These instances offer up to 60% better ML training performance and 20% lower cost compared to P4d instances, accelerating time to market for large dataset training. This enhancement benefits ML engineers and architects working with high-resolution data.
feature announcement - AWS What's New aiawsengineer ·
Amazon Bedrock adds Service Quotas support for bedrock-mantle endpoint
Amazon Bedrock now exposes inference quotas for its bedrock-mantle endpoint via AWS Service Quotas, providing customers with a unified view of their limits. This enhancement enables better planning for generative AI workloads by offering visibility into input and output token limits for supported models. Customers can access these quotas through the AWS Service Quotas console in all supported regions where the endpoint is available.
feature - AWS What's New aiawsengineer ·
AWS Elemental Inference Adds AI-Powered Live Subtitles
AWS Elemental Inference now offers AI-powered "Smart Subtitles" for live video streams, automatically generating real-time captions in TTML format across multiple languages. This feature aims to improve accessibility and reduce the need for manual captioning or third-party services for broadcasters and streamers. Users can enhance transcription accuracy via custom dictionaries and benefit from existing integrations with AWS Elemental MediaLive and other Elemental Inference capabilities.
feature - AWS What's New aiawsengineer ·
SageMaker Notebook Instances Add P5.4xl Instance Support
Amazon SageMaker notebook instances now support EC2 P5.4xl instances, featuring NVIDIA H100 GPUs for accelerated deep learning and HPC workloads. This enhancement can improve training times by up to 4x and reduce costs by up to 40%, benefiting engineers and data scientists working with large language models and generative AI applications. P5 instances are available in select AWS regions, with setup instructions provided in developer guides.
feature - AWS What's New aiawsengineer ·
SageMaker Notebook Instances add P5en.48xl instance types
Amazon SageMaker notebook instances now support the new EC2 P5en.48xl instance types, featuring H200 GPUs with enhanced memory and bandwidth. This upgrade significantly boosts AI training and inference performance, particularly for distributed workloads like deep learning and generative AI. These instances are now available in select AWS regions and are designed for users running demanding AI and HPC applications.
feature - AWS What's New aiawsengineer ·
Amazon Connect uses generative AI for automated self-service interaction evaluation
Amazon Connect now integrates generative AI to automatically evaluate self-service customer interactions, providing managers with aggregated insights for improving customer experience. Users can define custom evaluation criteria in natural language, and the AI will assess interaction quality, offering detailed reasoning. This feature is available in select AWS regions and aims to help identify opportunities for enhancing AI agent performance.
feature - AWS What's New aiawsengineer ·
SageMaker HyperPod adds MinCount for Slurm clusters
Amazon SageMaker HyperPod now allows users to specify minimum capacity requirements (MinCount) for Slurm-orchestrated clusters using continuous provisioning. This enhancement ensures distributed AI/ML training jobs start with a guaranteed number of nodes, preventing issues with partial cluster capacity. This feature is available in all AWS Regions where SageMaker HyperPod is supported and is particularly beneficial for large-scale distributed training.
feature - AWS What's New aiawsengineer ·
Amazon Aurora MySQL integrates with Kiro Powers for AI-assisted development
Amazon Aurora MySQL now integrates with Kiro Powers, a repository of AI agent tools, to help developers build applications faster. This integration offers conversational control over database operations and configuration, reducing the need for complex syntax. It provides task-specific guidance for scaling, migration, and replication, and is available via one-click installation in all AWS Regions where Aurora MySQL is supported.
feature announcement - AWS What's New aiawsengineer ·
SageMaker Unified Studio expands domain management for Identity Center domains
Amazon SageMaker Unified Studio now offers domain management for Identity Center-based domains outside the AWS console, enabling administrators to manage projects, workforce identity, and networking. This expands capabilities previously limited to IAM-based domains, improving unified administration. These features are available in all regions where SageMaker Unified Studio is supported.
feature - AWS What's New aiawsengineermedia ·
SageMaker Inference Supports OpenAI-Compatible APIs
Amazon SageMaker Inference now supports OpenAI-compatible APIs, allowing direct integration with tools like OpenAI SDK and LangChain by simply changing the endpoint URL. This feature simplifies connecting to SageMaker endpoints, enabling users to leverage existing code and authentication with custom models and VPCs. It offers flexibility in instance choice, data privacy, model execution, and autoscaling, with authentication managed via AWS credentials. The capability is available today in multiple AWS regions.
feature announcement - AWS What's New aiawsgaengineer ·
Amazon Bedrock adds request-level usage attribution
Amazon Bedrock now supports request-level usage attribution for InvokeModel and InvokeModelWithResponseStream APIs, allowing granular tracking of inference usage across teams and applications. This enhancement provides deeper visibility into consumption patterns, aids cost optimization, and simplifies internal reporting without requiring new resources. The feature is available in all Amazon Bedrock commercial regions and builds upon existing attribution capabilities.
feature - AWS What's New aiawsengineer ·
SageMaker HyperPod adds data capture for inference workloads
Amazon SageMaker HyperPod now supports data capture for inference workloads, recording request and response payloads to Amazon S3. This feature provides crucial visibility for generative AI deployments, enabling drift detection, troubleshooting, and model improvement without custom logging pipelines. It is available for SageMaker HyperPod clusters using the EKS orchestrator in all supported AWS Regions.
feature - AWS What's New aiawsengineer ·
SageMaker Studio IDEs support GPU capacity reservation
Amazon SageMaker Studio IDEs now support GPU capacity reservations via SageMaker Flexible Training Plans (FTP). This provides predictable access to high-performance GPU resources with potential cost savings of up to 65% compared to On-Demand instances. The feature is available for users running ML workflows in JupyterLab or Code Editor within Studio, offering a self-serve procurement experience.
feature