OpenAI's GPT-5.5 and Codex now generally available on Amazon Bedrock

OpenAI has expanded its partnership with AWS, making GPT-5.5, GPT-5.4, and the Codex coding agent generally available on Amazon Bedrock. The frontier models run on Bedrock's next-generation inference engine, which offers isolated queues for predictable performance under load. A key differentiator is durable state capture: if a hardware failure occurs mid-request, the execution resumes from the last checkpoint instead of restarting. Pricing matches OpenAI's first-party rates per token with zero additional fees. GPT-5.5 is the most capable model, grasping intent faster and autonomously handling complex multi-step tasks across codebases, document generation, data analysis, and tool orchestration.

Codex, the AI coding agent used by over 5 million developers weekly, is now available on Bedrock with pay-per-token pricing and no seat licenses. It holds repository-wide context, debugs ambiguous failures, and propagates changes across interconnected systems. GPT-5.5 powers Codex's inference for higher efficiency and quality. Deployment options include the Codex App, CLI, and IDE integrations (VS Code, JetBrains, Xcode). Enterprises benefit from full AWS governance: IAM roles, VPC isolation, KMS encryption, CloudTrail logging, and regional data residency. Prompts and responses remain private and are not used to train OpenAI models. This launch gives enterprises a unified platform to scale frontier AI with the security and compliance of AWS.