Amazon Bedrock customers can now invoke several OpenAI model families from the Canada Central and Canada West regions through cross-region inference. Available options include GPT-5.6, GPT-6, and GPT-6.1 variants.
Applications call the Bedrock Runtime endpoint in their Canadian source region. Bedrock then selects a supported destination region with available capacity, removing the need for the application to implement routing. The models can be used through OpenAI-compatible Responses and Chat Completions interfaces or Amazon’s Converse API.
Two routing profiles create different data-location boundaries. A US geographic profile keeps processing within Canada and the United States, while a global profile can use supported commercial AWS regions elsewhere. Availability varies by model and source region, and AWS notes that profile membership can change.
The models support text and image input, text output, tool calling, reasoning modes, prompt caching, and context windows of up to one million tokens, with model-specific differences. Organizations with strict residency rules should inspect the selected profile rather than assume that calling a Canadian endpoint means processing remains in Canada.