AWS Recaps September Bedrock Updates: Managed Agents Enter Preview Alongside New Models and Sync Controls

AWS published its September update recap on October 9, covering Amazon Bedrock, AgentCore, and Strands. Amazon Bedrock Managed Agents, powered by OpenAI, entered public preview. AWS says the agents can use OpenAI models while keeping data within AWS, reusing existing IAM permissions, and providing auditability through CloudTrail. The latest AgentCore runtime also brings more efficient memory management and lower cold-start latency; idle sessions scale to zero, with pay-as-you-go pricing and no pre-provisioned capacity.

For agent development, AWS says the open-source Strands harness matches popular harnesses on accuracy while using 28% fewer tokens, and supports Python or TypeScript. Strands Decider 2B is an open-source, 2-billion-parameter decision model that returns answers locally in about 115 milliseconds. These performance descriptions are claims in AWS’s recap.

The model updates include the general availability of Astra, Sol, and Luna on Amazon Bedrock. GPT-6 Astra supports up to 1 million input tokens, while GPT-6 Astra Ultrafast offers up to six-times-faster inference in the API, reaching up to 300 tokens per second. AWS also lists GPT-6.1 Sol and GPT-6 as available on Bedrock, and positions GPT 6.1 Luna for high-volume extraction, summarization, classification, and routing. OpenAI had introduced GPT-6.1 Sol on September 29. AWS further says Claude Sonnet 5.5 costs 30% less per task and runs 30% faster than Claude Sonnet 5. Moonshot AI’s Kimi K3 is available on Bedrock with a 1-million-token context window and native vision; xAI’s Grok 4.6 and Grok 4.7 offer a 500,000-token context window.

For enterprise data workflows, Amazon Bedrock Managed Knowledge Base added daily, weekly, and monthly automatic sync scheduling for all native data-source connectors. A user-managed setup for SharePoint, OneDrive, and Confluence also reduces reliance on admin-managed service accounts.