Introducing Gemma 4 models on Amazon Bedrock
TL;DR
AWS is bringing the Gemma 4 family to Amazon Bedrock. The models are built by Google DeepMind, released as open-weight models under Apache 2.0, and pitched around high intelligence per parameter. Bedrock gets three instruction-tuned variants: Gemma 4 31B as a dense 30.7B model, Gemma 4 26B-A4B as a MoE model with 25.2B total and 3.8B active parameters, and Gemma 4 E2B with 2.3B effective parameters.
Nauti's Take
AWS is not just adding another model to the shelf; it is offering a convenient escape hatch from proprietary model silos. The real test for AI teams is whether Gemma 4 is strong enough to push long context, tool calls, and multimodality into existing Bedrock stacks at a better cost profile.
Briefingshow
For teams that want open models without running their own inference stack, this is a practical move. The interesting part is the mix of open-weight models, Bedrock security controls, long context windows, and an OpenAI-compatible API. The announcement is clearly AWS-shaped, but it gives concrete deployment options rather than just model marketing.
Sources
