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 aiinfraawsengineer ·
AWS MCP Server adds cross-account and cross-role access
The AWS Model Context Protocol (MCP) Server now supports accessing multiple AWS accounts and IAM roles within a single session, eliminating restarts. This enhancement allows AI coding agents to switch seamlessly between environments, enabling faster workflows for tasks like cross-account debugging or configuration management. The feature is particularly beneficial for DevOps engineers and application developers working in multi-account setups and is available in US East (N. Virginia) and Europe (Frankfurt) Regions.
feature - AWS What's New aiawsengineerfinance ·
SageMaker Data Agent integrates business context from SageMaker Catalog
Amazon SageMaker Data Agent now integrates with SageMaker Catalog's business context and metadata, allowing data practitioners to use business terms for data discovery and code generation. This feature leverages curated business context from sources like Collibra and Atlan to improve accuracy and efficiency. It supports natural language queries for data retrieval and code generation, respecting data governance policies. This integration is available in SageMaker Unified Studio notebooks and Query Editor across supported AWS Regions.
feature - AWS What's New aiawsengineer ·
Amazon Bedrock Console Redesigned for OpenAI, Anthropic APIs
Amazon Bedrock has launched a redesigned console experience optimized for the bedrock-mantle endpoint, which supports OpenAI and Anthropic-compatible APIs. This update streamlines the process of experimenting with, iterating on, and scaling generative AI applications by organizing work into projects and providing project-aware documentation. The new console is available in all AWS Regions offering the bedrock-mantle endpoint and is designed for engineers and architects building with foundation models.
feature - AWS What's New aiawsengineer ·
Amazon SageMaker Data Agent adds conversation history
Amazon SageMaker Data Agent now supports conversation history within SageMaker Unified Studio. This feature allows data practitioners to maintain continuity across analytical sessions, resuming multi-step analyses and referencing past interactions. Conversation history is available in all regions where the Data Agent is offered and is accessible via a clock icon in the chat panel.
feature - AWS What's New aiawsengineer ·
SageMaker Studio adds notebook scheduling and orchestration
Amazon SageMaker Unified Studio now allows scheduling, parameterizing, and orchestrating notebook runs directly from the interface, reducing the need for external orchestration tools. This feature simplifies moving notebooks from experimentation to production by automating recurring tasks like daily reports and model retraining. It supports background runs, parameterized notebooks for reuse, and multi-notebook workflows. AI-assisted troubleshooting is also integrated to help resolve failures.
feature - AWS What's New aiawspreviewengineer ·
AWS Step Functions integrates with Amazon Bedrock AgentCore for AI reasoning
AWS Step Functions now supports AI agent reasoning steps via integration with Amazon Bedrock AgentCore, currently in preview. This feature automates complex tasks like document classification or data extraction within workflows, allowing for parallel execution, human approval, and detailed audit trails. It's available in select AWS Regions, with standard pricing for both services applying.
feature announcement - AWS What's New aiawsgagovernment ·
OpenAI GPT-5.4 available on Amazon Bedrock in AWS GovCloud (US-West)
OpenAI's GPT-5.4 model is now generally available on Amazon Bedrock, specifically within the AWS GovCloud (US-West) region. This enables government and regulated industries to leverage advanced AI capabilities with enhanced security and compliance features. The model runs on a new, high-performance inference engine, ensuring data privacy by keeping prompts and responses within the customer's AWS environment.
feature announcement - AWS What's New aiawsengineer ·
Amazon SageMaker AI adds multi-turn reinforcement learning
Amazon SageMaker AI now supports multi-turn reinforcement learning for customizing foundation models on multi-step, agentic tasks. This serverless capability simplifies the complex process of training AI agents by rewarding full decision sequences, enabling specialization of smaller models for specific workloads. It is available today through SageMaker Studio and the Python SDK, with support for various models and AWS compute options.
feature - AWS What's New aiinfraawsengineer ·
Amazon ECS Managed Instances add support for AWS Trainium and Inferentia
Amazon ECS Managed Instances now supports AWS Trainium and AWS Inferentia, purpose-built AI accelerators for generative AI workloads. This feature simplifies infrastructure management for ECS users by offloading operations to AWS, enhancing performance and reducing costs. Users can now select accelerated instance types and configure resource requirements within their task definitions to allocate accelerators efficiently. This update is available immediately via the AWS Console, ECS MCP Server, or IaC tooling.
feature - AWS What's New aiawsengineerfinance ·
Amazon ElastiCache for Valkey adds durability support
AWS has announced durability support for Amazon ElastiCache for Valkey, allowing workloads requiring low latency and high availability to prevent data loss. Customers can choose between synchronous or asynchronous writes to meet different data loss tolerance and latency requirements. This enhancement expands ElastiCache's applicability to critical use cases like AI agent memory and payment processing, and is available globally starting with Valkey 9.0.
feature - AWS What's New aiawsengineer ·
Amazon SageMaker Studio setup faster, includes model customization permissions
Amazon SageMaker Studio now sets up in under twenty seconds, a significant reduction from previous setup times. This accelerated setup process automatically configures serverless model customization permissions, simplifying the process for ML engineers and data scientists. A new managed policy, AmazonSageMakerModelCustomizationCoreAccess, is now attached by default for new environments, reducing manual IAM configuration. This feature is available in all supported AWS Commercial Regions.
feature patch - AWS What's New aiawsgaengineer ·
Amazon Bedrock AgentCore Identity supports BYO secrets with Secrets Manager
Amazon Bedrock AgentCore Identity now allows referencing existing AWS Secrets Manager secrets directly, enhancing governance and control. Previously, secrets were service-managed, hindering custom encryption, tagging, and policy enforcement. This update enables customers to manage their secrets with their own policies and CMKs, providing full ownership and governance, and is now generally available across 14 AWS regions.
feature announcement - AWS What's New aiinfraawsengineer ·
SageMaker HyperPod adds EFA-only network interface option
Amazon SageMaker HyperPod now supports EFA-only network interfaces, allowing users to configure dedicated Elastic Fabric Adapters without requiring traditional Elastic Network Adapters for IP networking. This change is crucial for large-scale AI/ML model development, enabling further scaling of training clusters by preventing IP address exhaustion within VPCs. The feature is available in all AWS Regions where SageMaker HyperPod is supported and is configured via the CreateCluster/UpdateCluster API.
feature - AWS What's New aiawsengineer ·
Amazon Quick now supports private VPC connectivity for MCP servers
Amazon Quick now allows private connections to on-premises Model Context Protocol (MCP) servers via VPC, expanding AI workflow capabilities for enterprises. This feature enables secure integration with proprietary applications and internal tools hosted on private networks, previously limited to public internet access. It benefits organizations hosting MCP servers on EC2, Fargate, or Agentcore, allowing natural language interaction with private data sources through Quick, with all traffic secured within your VPC. This enhancement is available across all AWS Regions where Quick is offered.
feature - AWS What's New aiawsengineer ·
Amazon Bedrock adds CloudWatch metrics for OpenAI/Anthropic APIs
Amazon Bedrock now provides Amazon CloudWatch metrics for inference traffic to the bedrock-mantle endpoint, which supports OpenAI and Anthropic APIs. This allows customers to monitor usage, set up alarms, and plan capacity for generative AI applications running on Bedrock. The new metrics are available in all regions where the bedrock-mantle endpoint is offered and can be accessed via the AWS/BedrockMantle namespace in the CloudWatch console.
feature - AWS What's New aiawsgaengineer ·
OpenAI's GPT-5.5, GPT-5.4, and Codex now available on Amazon Bedrock
Amazon Bedrock now offers production access to OpenAI's GPT-5.5 and GPT-5.4 models, alongside Codex for AI-assisted software development. This integration provides enhanced capabilities for coding, data analysis, and autonomous tasks within the existing AWS security and governance framework. Pricing aligns with OpenAI's rates and counts towards AWS commitments, making these advanced AI models accessible for enterprise use.
feature announcement - AWS What's New aiawsengineer ·
SageMaker HyperPod adds AI coding assistant troubleshooting skills
Amazon SageMaker HyperPod now integrates troubleshooting skills with AI coding assistants like Claude Code, Cursor, and Kiro. This feature enables expert-level cluster diagnostics through natural language, streamlining the resolution of complex issues in large-scale AI/ML infrastructure. It benefits operators by reducing the time and expertise needed for debugging, without requiring modifications to existing HyperPod setups. The skills are available today as open source for Slurm and Amazon EKS orchestrated clusters.
feature announcement - AWS What's New aiawsengineerenergy ·
AWS PCS offers production-ready Deep Learning AMI
AWS Parallel Computing Service (AWS PCS) has launched a production-ready Deep Learning AMI (DLAMI) built on the Deep Learning Base GPU AMI. This AMI simplifies AI/ML training and HPC by pre-installing core infrastructure components like the PCS Agent and Slurm, making it easier to build and scale environments. It is available at no additional cost in all AWS Regions where PCS is available and supports both x86_64 and arm64 architectures.
feature - AWS What's New aiawsgaengineer ·
Amazon Connect adds generative AI post-contact summaries in 8 new languages
Amazon Connect now offers generative AI-powered post-contact summaries in eight new languages, including Portuguese, French, and German, along with regional English variations. This enhancement allows for automated, concise overviews of customer interactions across channels, reducing the need for agents to review full transcripts. Organizations can now generate summaries in the language of the conversation, improving agent efficiency and manager oversight for global support operations. These new language capabilities are available in all AWS Regions where the feature is currently supported.
feature - AWS What's New aiawsgaengineer ·
Claude Opus 4.8 available on AWS via Bedrock and Claude Platform
Anthropic's Claude Opus 4.8, their most capable general availability model, is now accessible on AWS. This model offers significant improvements in agentic coding, complex knowledge work, and long-running autonomous tasks, benefiting developers and enterprises building AI applications. Customers can access it through Amazon Bedrock for integrated AWS features or the Claude Platform on AWS for direct Anthropic experience with AWS billing and authentication.
feature announcement