AWS adds cross-Region inference for OpenAI GPT-5.6 models on Amazon Bedrock

Amazon Web Services has added cross-Region inference for OpenAI's GPT-5.6 model family on Amazon Bedrock, letting requests route across multiple AWS regions based on available capacity rather than being pinned to wherever an application is deployed.

What's new

According to the AWS Machine Learning blog, "Amazon Bedrock now offers OpenAI GPT-5.6 models on Amazon Bedrock in more than 25 AWS Regions, with cross-Region inference." The feature covers three GPT-5.6 variants — Sol, Terra, and Luna — each pitched at a different capability-to-cost tradeoff, and spans more than 25 regions across the US, Canada, Europe, Asia Pacific, the Middle East, and South America.

Developers get two ways to route requests: geographic inference profiles (for example, us.openai.gpt-5.6-terra) that keep processing confined to a specific geography for data-residency purposes, and global inference profiles that route a request across all supported regions to find available capacity. AWS says the cross-Region option is aimed at improving throughput and keeping performance consistent under load by drawing on a broader pool of compute rather than a single region's capacity. Pricing follows Bedrock's standard published rates rather than a new pricing tier.

Context

Bedrock added GPT-5.6 models to its OpenAI lineup earlier this year alongside Anthropic, Meta, and other third-party model providers, part of Amazon's strategy of positioning Bedrock as a model-agnostic hosting layer rather than a single-vendor API. Cross-Region inference itself isn't new to Bedrock — AWS has offered it for Anthropic and Amazon's own models before — so this release extends an existing routing mechanism to the OpenAI models Bedrock added most recently.

Why it matters

For enterprise customers already committed to Bedrock as their inference layer, this closes a capability gap between OpenAI's models and the rest of Bedrock's catalog: teams that need both data-residency guarantees and burst capacity for GPT-5.6 workloads no longer have to choose between a single-region deployment and a separate provisioning path. It's also a signal of how deep AWS's OpenAI hosting relationship has become — treating a competitor's flagship model family with the same regional-routing infrastructure AWS built for its own models.