AWS unveils SageMaker multi-turn RL, Studio speedup, and OpenAI GPT-5.5 on Bedrock

Amazon Web Services announced three major AI and machine learning updates on June 3, 2026. The releases include a new multi-turn reinforcement learning feature on SageMaker AI, a faster SageMaker Studio setup, and the general availability of OpenAI's GPT-5.5 and Codex on Amazon Bedrock.

SageMaker Studio now completes its quick setup in under twenty seconds, down from over two minutes. The update automatically configures serverless model customization permissions via a new managed policy called AmazonSageMakerModelCustomizationCoreAccess.

SageMaker AI Multi-Turn RL Targets Agentic Tasks

The multi-turn reinforcement learning offering on SageMaker AI lets customers fine-tune models on multi-step agentic tasks. It is a serverless capability: users pay only for tokens processed, with no infrastructure to provision. The service manages the full training loop, including rollout orchestration and trajectory collection. Built-in MLflow tracking allows inspection of agent trajectories and rewards.

Supported models include Qwen 3.6 27B, Nova Lite 2.0, GPT-OSS-20B, and Gemma 31B in us-west-2, and Nova Lite 2.0 and GPT-OSS-20B in us-east-1. Customers can connect their agent running on Amazon Bedrock AgentCore Runtime, Amazon EKS, Amazon EC2, AWS Fargate, or custom infrastructure.

• Multi-turn RL trains models by rewarding the full sequence of decisions across a task.
• It helps specialize smaller models to match larger ones on target workloads.
• Evaluation jobs report reward, pass@k, and trajectory metrics.
• Available through SageMaker Studio and the SageMaker Python SDK.

OpenAI's GPT-5.5 and Codex Land on Bedrock

OpenAI's GPT-5.5 and GPT-5.4 models are now generally available on Amazon Bedrock for production workloads. GPT-5.5 is described as the most capable OpenAI model, excelling at agentic coding, data analysis, and multi-step autonomous tasks. It runs on the Bedrock next-generation inference engine, built for high performance and security. Codex, an AI-powered software development tool, is available through the Codex App, CLI, and IDE integrations with Visual Studio Code, JetBrains, and Xcode.

The suite of updates signals AWS's continued investment in agentic AI and rapid developer onboarding. With SageMaker Studio launching in under 20 seconds and serverless model customization permissions pre-configured, AWS is reducing friction for machine learning practitioners. Meanwhile, the addition of frontier OpenAI models on Bedrock gives enterprise customers access to advanced capabilities under existing AWS governance controls.

What comes next: AWS is expected to expand the multi-turn RL capability to additional regions and models. The company also hinted at deeper integrations between SageMaker and Bedrock agent runtimes in upcoming releases.