AWS adds Claude Sonnet 5 as it trims older services
Mon, 6th Jul 2026 (Today)
Amazon Web Services has added a series of artificial intelligence and infrastructure products and updated the availability status of a wide range of existing services. The changes span model access, desktop automation, observability tools, and product lifecycle notices.
New additions include Claude Sonnet 5 on AWS, general availability for Amazon WorkSpaces for AI agents, a new log analytics engine in Amazon OpenSearch Service, faster scale-out for generative AI inference in Amazon SageMaker AI, and a way to create Amazon CloudWatch alarms from log queries.
Taken together, the updates reflect AWS's current focus on AI software, developer tools, and operational management, while also showing a broad cleanup of older or less strategic products across identity, search, desktop computing, management services, and parts of SageMaker AI.
AI additions
Among the headline launches, Claude Sonnet 5 is now available on AWS. The Anthropic model is aimed at coding, agents, and general professional tasks, and can work across large codebases, call tools, and maintain state over longer-running agent tasks.
Amazon WorkSpaces for AI agents is also now generally available. The service is designed to let AI agents access and operate desktop applications through managed WorkSpaces environments, allowing customers to use existing desktop software without rebuilding those applications or adding custom integrations.
The release is notable because it targets a problem many large organisations face when trying to apply AI systems to long-established business software. Rather than requiring direct application programming interfaces or a full software overhaul, the approach centres on desktop access in a managed environment.
Amazon SageMaker AI also received an inference scaling update. SageMaker Inference now supports container image caching, which can cut end-to-end scale-out time for generative AI models by as much as half when extra capacity is needed.
Operations focus
On the infrastructure and operations side, AWS introduced a new engine in Amazon OpenSearch Service for log analytics workloads. The engine is built for log analysis while retaining full-text search, allowing teams to run aggregations and text search in the same system.
AWS says the new OpenSearch engine delivers up to four times better price-performance on its internal benchmarks. The claim will be watched closely by customers facing rising observability costs as application estates grow and machine-generated data volumes increase.
CloudWatch also gained a feature that allows alarms to be created directly from log query results. This removes the need to create metric filters or custom metrics first, reducing the steps needed to trigger alerts from log data.
Another infrastructure update came in AWS CloudFormation, where a new Express mode is intended to shorten infrastructure deployment times. The feature is available in all commercial regions at no extra cost and is designed to return deployment confirmation in seconds.
For container users, Amazon EKS now includes Kubernetes version rollbacks within seven days of an upgrade. That allows customers to reverse cluster upgrades if problems emerge without rebuilding the cluster.
AWS also released Amazon EC2 C9g and C9gd instances based on its Graviton5 processors. The instances offer higher compute performance than Graviton4-based versions, a larger cache, and, in the C9gd option, local NVMe storage.
Product changes
Alongside the launches, AWS updated the status of a long list of services in its product lifecycle notices. Several products are moving into maintenance status, meaning they will no longer be accessible to new customers after the relevant cut-off dates, while others are entering sunset or have reached end of support.
Products moving to maintenance include Amazon Bedrock Agents, now referred to as Amazon Bedrock Agents Classic; Amazon Cognito Sync; Amazon Kendra; Amazon Q Business; AWS Directory Service Simple AD; AWS Mainframe Modernization Self-Managed Experience; AWS Management Console myApplications; AWS Resource Groups Group Lifecycle Events; AWS Service Catalog Application Registry; and AWS Systems Manager Application Manager.
AWS IoT Device Defender Detect is also moving to maintenance, with a separate access deadline for new customers. A broad set of SageMaker AI features has been placed in the same category, including A2I, Clarify, Debugger, GeoSpatial, Ground Truth, Mechanical Turk, Model Monitor, Role Manager, and Studio Lab.
Services entering sunset include Amazon WorkSpaces PCoIP, Amazon WorkSpaces Pool, AWS Managed Services Advanced, AWS re:Post Private, and Amazon SageMaker AI Profiler.
Amazon Chime SDK Carrier Voice Focus and Amazon SageMaker AI Ground Truth Plus have reached end of support.
The lifecycle changes are significant because they affect both new buying decisions and existing migration plans. For customers with large AWS estates, especially those that adopted specialist machine learning and management tools over several years, the notices may prompt reviews of product roadmaps, replacement options, and internal support timelines.
The breadth of the changes also suggests AWS is refining its portfolio as it pushes further into AI platforms and newer infrastructure services, while drawing down older offerings whose roles may now overlap with newer products or partner tools.
Separately, Daniel Abib, who wrote the weekly AWS update, pointed to startup activity as an example of how cloud and AI services are being used to solve specific problems. He highlighted EyeCare Health, a Brazilian health technology company, which is using smartphones to widen access to basic eye screening in places where specialist care is scarce.
He wrote that more than 70% of Brazilian municipalities do not have a single ophthalmologist, underscoring the kind of healthcare access gap some startups are trying to address with software delivered through widely used consumer devices.