AWS brings open OpenAI and NVIDIA models into Bedrock GovCloud

AWS published a new Bedrock update for regulated environments on April 29: OpenAI GPT OSS models and NVIDIA Nemotron models are now available in Amazon Bedrock in AWS GovCloud (US).

What is new

The update covers OpenAI GPT OSS 120B and 20B, plus NVIDIA Nemotron Nano 9B v2, Nano 12B v2, Nano 30B and Super 120B. AWS describes the models as open, with emphasis on transparency, flexibility and specialized agentic AI systems.

AWS also says the models are powered by Mantle, a new distributed inference engine for Bedrock. Mantle is meant to speed up model onboarding, provide serverless inference, higher default quotas, capacity management and compatibility with OpenAI API specifications.

Why it matters

This is quieter than GPT-5.5 arriving on Bedrock, but it matters for regulated organizations. GovCloud customers get more open model options behind one API surface, with AWS security, scaling and compliance controls around them.

For CIOs, it makes model strategy more modular. Proprietary frontier models, open models and specialized agent models can sit inside the same operational and control model. That is useful for public sector, defense, healthcare, finance and other environments where data locality, auditability and vendor control matter.

Source and date validation

The original source is AWS What's New, published on April 29, 2026: https://aws.amazon.com/about-aws/whats-new/2026/04/openai-gpt-oss-nvidia-nemotron-govcloud/. The item is official and within the 48-hour freshness window.