GPT 5.5 lands on Bedrock for production AI

GPT 5.5 just hit production on Bedrock, and pricing matches OpenAI direct.

Amazon Bedrock now hosts GPT-5.5, GPT-5.4, and Codex as generally available, the company announced. The models run on Bedrock’s next generation inference engine, a setup AWS says is built for high performance, reliability, and security in production. Pricing is the lever Bedrock emphasizes: GPT 5.5 and 5.4 are offered at the same per token rate as OpenAI first party, and Codex is available with pay per token pricing. Inference runs through Bedrock, and usage counts count toward existing AWS commitments. One month after the expanded partnership announcement, these frontier models are ready to deploy in production applications and agents via Bedrock’s Responses API.

The team reports GPT 5.5 is the most capable OpenAI frontier model on Bedrock, designed to grasp intent faster and handle multi step tasks autonomously. In practical terms, it excels at writing and debugging code across large code bases, analyzing data, generating documents and spreadsheets, and operating software across multiple tools until a task is complete. The improvements are most notable in agentic coding and knowledge work, where progress depends on sustaining context and taking action over time. GPT 5.4 joins GPT 5.5 in the catalog, and Codex remains a focal point for software development workflows on Bedrock.

Bedrock emphasizes not just the models themselves but the platform that runs them. The inference engine provides an isolated queue with automated capacity management, aiming to smooth out production load and reduce surprising latency as teams scale. For developers, that means you can run frontier models with predictable throughput and without building your own tooling to manage scale, at least within the Bedrock ecosystem. The Models sit in Bedrock’s catalog and are accessed through the Responses API, with the same per token economics as the direct OpenAI options and no hidden Bedrock fees.

For practitioners, the availability shifts how teams think about embedding AI into product workflows. First, enterprise buyers can bind these models to existing AWS commitments, turning AI usage into a known line item rather than an ad hoc experiment. Second, coders and developer advocates gain a turnkey path to build agent-like pipelines that write, test, and operate software across multiple tools in a single workflow. Third, product managers have a clearer cost and reliability envelope, since Bedrock handles capacity and isolation without forcing a tiered mix of services.

Two concrete insights for engineers and leaders watching this space:

The broader takeaway is clear: production-grade frontier AI is moving closer to everyday software pipelines on major clouds. Bedrock’s generally available push for GPT 5.5, 5.4, and Codex tightens the linkage between AI capability and enterprise operations, not just experimentation. As teams begin to ship autonomous coding and data tasks at scale, the next set of watchpoints will be latency under complex toolchains, the efficiency of long-context sessions, and the governance controls that keep such agents aligned with business needs.