ICYMI: What landed for AI builders in September 2026
A monthly recap of the latest Amazon Bedrock, Amazon Bedrock AgentCore, and Strands updates from September 2026: broader model choice, faster serverless agents with built-in evaluation, and automated knowledge base syncing with native enterprise connectors.
In September 2026, Amazon unveiled several updates to its AI offerings aimed at providing builders greater choice and efficiency across model, runtime, and tooling layers. These updates include Amazon Bedrock, AgentCore, and Strands.
Amazon Bedrock continues to expand its selection of leading frontier models, enabling the secure building and scaling of AI applications. The platform now offers Amazon Bedrock Managed Agents, powered by OpenAI's models, in public preview. This allows developers to create production-ready AI agents without compromising enterprise security while reusing AWS IAM permissions and maintaining full auditability through AWS CloudTrail.
The latest AgentCore runtime provides more efficient memory management and lower cold start latency for serverless agents. This change results in reduced infrastructure overhead and lower costs, with agents now scaling to zero when idle and running in hardware-isolated environments.
Strands, an open-source toolkit from Amazon for building agents, is also updated. Strands harness now uses 28% fewer tokens than popular alternatives while maintaining accuracy. This toolkit simplifies the creation of production-ready agents with built-in context management, prompt caching, and memory, deployable in a single line of Python or TypeScript.
In terms of model options, Amazon Bedrock now supports OpenAI's Astra, Sol, and Luna models, including GPT-6 Astra Ultrafast for faster inference and GPT-6 Sol for coding and professional workloads. Claude models have been updated to include Claude Fable 5.1 for coding, scientific research, and enterprise workflows, Claude Opus 5.5 for agentic coding and long-running tasks, and Claude Sonnet 5.5 for smarter, more efficient coding and knowledge work.
Additionally, Moonshot AI's Kimi K3, its most capable model with a 2.8 trillion parameter count and 1-million-token context window, and xAI's Grok 4.6 and Grok 4.7, with a 500K token context window and configurable reasoning effort, are now available on Amazon Bedrock. These additions provide builders with more flexibility in matching models to their specific requirements.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
