AWS AI Updates: November 18, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. OpenAI GPT-6.2 Astra and GPT-6.2 Sol Reach General Availability on Amazon Bedrock
Amazon Bedrock now offers two OpenAI models, GPT-6.2 Astra and GPT-6.2 Sol, as generally available. Astra targets the hardest work, including complex business problems, advanced coding and agentic workflows, and includes an Ultrafast mode for latency-sensitive uses such as interactive coding assistants and real-time agents. Sol is aimed at recurring, high-volume work where speed and cost per task matter, such as software implementation, debugging and multistep workflow automation. Both are available in the Bedrock console and through supported Bedrock APIs, with existing AWS controls covering access governance and invocation auditing. Source
2. Bedrock AgentCore Memory Raises Short-Term Event Retention From One Year to 10 Years
Amazon Bedrock AgentCore Memory now lets short-term memory events be retained for up to 10 years, up from one year. Retention is set on the memory resource and applied to each event when it is created, and existing events keep the retention they were created with. Standard short-term memory storage charges apply for as long as data is kept. The feature is available in every Region where AgentCore Memory is offered. Source
3. Bedrock AgentCore Memory Adds Full Session Summaries That Update as Events Arrive
Amazon Bedrock AgentCore Memory now keeps a single full session summary record for each session, alongside the existing summary chunks that cover specific parts of a conversation. The record updates as new events arrive. Any memory that uses a summary strategy gets it automatically, including for existing sessions, with no extra configuration. Summaries are stored as long-term memory records and can be retrieved with the existing Memory operations by filtering on record type. Source
4. Bedrock AgentCore Includes 5,000 Free Web Search Queries Per Month
Amazon Bedrock AgentCore now includes the first 5,000 web search queries each month at no charge, through the Web Search Tool in AgentCore Gateway and the Web Search API. The allowance applies automatically, is shared across supported Regions and across every account in an AWS organization, and resets monthly. Queries beyond 5,000 are billed at standard rates and agents keep running. Supported Regions are US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Sydney) and Asia Pacific (Tokyo). Source
5. SageMaker AI Adds Data Preparation Tools for Synthetic Data, Auto-Labeling and Quality Evaluation
Amazon SageMaker now includes AI data preparation capabilities for building training and evaluation datasets for text-based model customization and traditional ML. The four capabilities are synthetic data generation grounded in a team’s own documents, auto-labeling with a configurable confidence threshold that routes low-confidence rows to human review in Label Studio Community Edition, data quality evaluation for duplicates, tone and diversity issues, and data transformation with operators such as dedupe, filter and format conversion. They are available through the SageMaker Python SDK or through natural language with the AWS AI-ML skill in the Agent Toolkit for AWS, in US East (N. Virginia) and US West (Oregon). Data and labels stay in the customer’s S3 buckets. Source
6. SageMaker HyperPod Tracks Scaling and Deep Health Check Requests as Cluster Operations
Amazon SageMaker HyperPod now records each scaling request and on-demand Deep Health Check as a cluster operation, followed from submission through completion. Each operation shows what it changed in every instance group and has its related events attached, and when several scaling requests target the same instance group under continuous provisioning, it shows whether it completed or was superseded by a newer request. New ListClusterOperations and DescribeClusterOperation APIs support dashboards, alerting and CI/CD integration, and ListClusterEvents can filter by operation. Operations are created automatically at no additional cost in all Regions where HyperPod is available. Source
7. EMR Serverless Gets an Apache Spark Optimization Agent for Cost and Performance Recommendations
Amazon EMR Serverless introduced an Apache Spark optimization agent that analyzes completed job runs and returns cost or performance recommendations. A user picks a cost or performance goal, and a deterministic recommendation engine checks job configuration, logs and Spark telemetry for bottlenecks such as insufficient parallelism, skewed join keys, memory spill and idle executors. Each recommendation carries telemetry evidence and is tagged by layer: infrastructure, Spark configuration, code or data layout. It can also compare a run against a known-good baseline, and is reached through the EMR Serverless console or the SageMaker Unified Studio MCP Server. It requires EMR release 7.6 or later and has no additional charge. Source
8. AWS Lets Agentic Coding Harnesses Open Account Sign-Up and Receive Short-Lived Credentials
AWS announced agentic account creation, which lets coding harnesses such as Claude Code, Codex and Cursor open the AWS sign-up flow for developers who do not yet have an account. The browser opens automatically, the developer completes sign-up, and short-lived credentials are returned to the harness. Each new project starts with a home Region, an IAM role manager and credentials that refresh automatically. Developers need to install the AWS plugin or configure the AWS Agent Toolkit in their harness. Source