AWS now offers Google DeepMind's Gemma 4 open-weight AI models through its managed Bedrock service, giving businesses access to cutting-edge models while keeping their data fully under their own control.

What happened: AWS announced the availability of three Gemma 4 model variants on Amazon Bedrock, built by Google DeepMind and released under the Apache 2.0 license. The family spans from a 2.3B-effective-parameter compact model up to a 30.7B-parameter dense model, each supporting built-in reasoning, native function calling, and multimodal input (text and image).

Why it matters: Organizations adopting open-weight foundation models (AI systems built with publicly available code and weights) face a tension between using leading models and maintaining data protection and regulatory control. Amazon Bedrock removes that trade-off by running inference entirely on AWS infrastructure, ensuring prompts and completions are not used to train models and content is not shared with third parties.

What to watch: Independent benchmarks show Gemma 4 31B achieves an Intelligence Index of 39, well above the median of 15 in the 4B–40B open-weights class. The Gemma 4 26B-A4B variant uses a mixture-of-experts architecture with only 3.8B active parameters per request, delivering roughly 4B-class cost and latency while retaining the knowledge capacity of a much larger model.