Official AWS What's New (AI-filtered) & AWS ML Blog RSS + GitHub + Hugging Face.
39 updates · page 1 of 3
Amazon Bedrock now accepts the reasoning.summary parameter for OpenAI models through the Responses API, so you can ask for a readable summary of the model's reasoning next to its answer.
Why it matters: Developers debugging coding, analysis, or multi-step tasks can see how the model approached the problem.
Claude Sonnet 5.5 and Claude Opus 5.5 are now in the Kiro IDE and CLI for AWS GovCloud (US). Opus 5.5 thinks adaptively and, in Kiro's internal tests, uses about 40% fewer tool calls and about half the tokens of Opus 5.
Why it matters: GovCloud developers can use Opus 5.5 in Kiro for long-running agentic coding with fewer tool calls and tokens than Opus 5.
AWS recapped September 2026 updates across Amazon Bedrock, Amazon Bedrock AgentCore, and Strands, focused on model, runtime, and tooling choice for builders.
Why it matters: Builders can scan one recap to see which Bedrock, AgentCore, and Strands pieces landed in September 2026.
Building an AI agent for a demo and operating one for 40 million developers are different engineering problems.
Why it matters: Developer tooling release — update for the latest fixes, features, and compatibility.
Amazon SageMaker Unified Studio now supports custom Tooling blueprints, so domain administrators can define each project's foundation with their own AWS CloudFormation templates.
Why it matters: Admins can enforce naming standards and custom permission boundaries on every SageMaker project.
Amazon Bedrock now offers TwelveLabs Pegasus 1.5, a video-to-text model that reads a whole video and writes structured, time-coded metadata for what happens and when.
Why it matters: You can query a large video library by moment without already knowing where to look.
Ultrafast mode for OpenAI GPT-6.1 Sol is now on Amazon Bedrock, offering faster inference for latency-sensitive work such as real-time coding assistants.
Why it matters: Bedrock users can run GPT-6.1 Sol Ultrafast when a coding assistant or other interactive app needs lower latency.
AWS Cost Explorer, Budgets, and Cost Management Dashboards can now break Amazon Bedrock cost down by model, model provider, inference type, and feature.
Why it matters: Finance and platform teams can group, filter, and budget Bedrock spend by model and feature.
When an AI agent runs, it often needs to buy something to finish a task: a model inference, an API response, access web content, or a call to another agent.
Why it matters: Developer tooling release — update for the latest fixes, features, and compatibility.
Amazon shipped Strands Agents SDK harness-cli v0.2.0 (minor release) on GitHub.
Why it matters: New version with notable changes — read the release notes for breaking changes before upgrading.
Amazon shipped Strands Agents SDK harness-python v0.2.0 (minor release) on GitHub.
Why it matters: New version with notable changes — read the release notes for breaking changes before upgrading.
Amazon shipped Strands Agents SDK harness-typescript v0.2.0 (minor release) on GitHub.
Why it matters: New version with notable changes — read the release notes for breaking changes before upgrading.
Claude Haiku 5.5 is available on Amazon Bedrock and Claude Platform on AWS. Anthropic calls it the fastest and most efficient Claude 5.5 model, built for subagents and high-volume work, at around 75 percent less than Claude Haiku 4.5 for most tasks.
Why it matters: AWS teams can run high-volume subagent work on Haiku 5.5 while paying about 75 percent less than Haiku 4.5 for most tasks.
An AWS post rethinks access control for retrieval-augmented generation with Amazon Quick and Amazon Bedrock so answers respect permissions in sources such as SharePoint, Google Drive, and Confluence.
Why it matters: Teams indexing company files can keep generated answers inside the permissions those sources already enforce.
AWS argues that hours-saved times labor cost, a model built for rule-based RPA, misses most of the value of agentic automation, and introduces a framework for AI CoE leaders.
Why it matters: AI CoE leaders can use the framework when a simple hours-saved case understates what agents do.