AWS AI Updates: August 4, 2026

1. OpenAI GPT-5.6 models gain 1 million token context on Amazon Bedrock

AWS. OpenAI’s GPT-5.6 Sol, Terra, and Luna models now support 1 million token context windows on Amazon Bedrock, enabling single-request analysis of entire codebases, long legal or regulatory documents, and full conversation histories in multi-step agentic workflows. Sol is available in US East (N. Virginia and Ohio), while Terra and Luna add US West (Oregon), accessed through the Bedrock Console or the Responses API on the bedrock-mantle endpoint. Prompt caching with explicit cache breakpoints applies to long-context requests, so repeated context is billed at a caching discount. Source

2. Amazon SageMaker serverless customization adds full fine-tuning

AWS. Amazon SageMaker AI serverless model customization now supports full fine-tuning, updating all model parameters for deeper adaptation than parameter-efficient methods such as LoRA. The capability covers more than 25 open-source models across the gpt-oss, Gemma, Llama, Nemotron, and Qwen families, with SageMaker managing all infrastructure provisioning and training orchestration under consumption-based pricing. It targets cases requiring specialized reasoning patterns, complex output formats, or internalizing domain knowledge from large proprietary datasets, and is available in US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Ireland) via SageMaker Studio JumpStart or the Python SDK. Source

3. AWS Transform continuous modernization reaches general availability

AWS. AWS Transform continuous modernization is now generally available across all supported regions, letting engineering teams analyze and remediate technical debt across source code repositories at scale. It connects to GitHub organizations, GitLab groups, and Bitbucket workspaces, running on-demand or scheduled analysis that prioritizes findings across technical debt, security, agentic readiness, and modernization readiness. For remediable findings the system creates branches and opens pull requests or merge requests with validated code changes, with analysis and remediation running in the customer’s own AWS account. Source

4. AWS Config adds resource types for Bedrock and SageMaker governance

AWS. AWS Config now supports 15 new resource types, including several for Amazon Bedrock and SageMaker that expand governance over AI/ML workloads. New Bedrock types cover automated reasoning policies and versions, blueprints, data automation projects, and an Agent Core API key credential provider, while new SageMaker types include image versions, inference components, partner apps, projects, and spaces. Organizations with all resource types enabled will automatically begin tracking these additions in regions where the services operate. Source