AWS Bedrock Now Supports Open Weight Models in European Sovereign Cloud

European organizations can now run generative AI workloads using open-weight models on Amazon Bedrock in the AWS European Sovereign Cloud, with Gemma 4 being the first available model family. This allows for building, iterating, and scaling generative AI applications while meeting stringent digital sovereignty requirements by keeping data within the EU.

Gemma 4 is available under the Apache 2.0 license, offered in three instruction-tuned variants (31B, 26B-A4B, E2B) to suit different use cases like reasoning-heavy, cost-sensitive, or latency-sensitive workloads. These models provide built-in reasoning, native function calling, and multimodal input capabilities across text and image.

The Bedrock next-generation inference engine, accessible via the bedrock-mantle endpoint, is built for high performance, reliability, and security. It supports OpenAI-compatible APIs, allowing existing OpenAI SDK codebases to be adapted with minimal changes, and offers stateful conversation management.

Amazon Bedrock employs a zero operator access and zero data retention model, ensuring prompts and responses are private and not stored by default. All inference remains within the eusc-de-east-1 Region of the AWS European Sovereign Cloud, which is independently operated by EU-resident AWS employees to guarantee EU data residency and operational controls.