Run NVIDIA Nemotron and OpenAI GPT OSS models on Amazon Bedrock in AWS GovCloud (US)

2026-07-01 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Cybersecurity & Data Privacy · Depth: Intermediate, long

Summary

Amazon Bedrock in AWS GovCloud (US) now supports OpenAI's open-weight GPT OSS models (120B and 20B) and NVIDIA Nemotron models (Nano 9B v2, Nano 12B v2, Nano 30B, Super 120B), enabling U.S. government agencies to deploy advanced generative AI within strict security and compliance boundaries. This expansion allows agencies to run workloads like intelligence analysis and contract review on models such as the NVIDIA Nemotron 3 Super 120B, which features 120 billion total parameters and a 1-million-token context window, or the GPT OSS 120B with its 128K-token context window. Inference runs entirely within the AWS GovCloud (US) isolation boundary, administered by U.S. citizens, meeting frameworks like FedRAMP High and DoD SRG Impact Levels 2, 4, and 5. Users can access these models via "bedrock-mantle" (OpenAI-compatible API) or "bedrock-runtime" (AWS SDK), with options for In-Region or Geo Cross-Region inference across "us-gov-west-1" and "us-gov-east-1".

Key takeaway

For AI Engineers and MLOps Engineers deploying generative AI in U.S. government or defense contexts, you can now securely integrate advanced open-weight models like OpenAI GPT OSS and NVIDIA Nemotron directly within AWS GovCloud (US). This eliminates the need to compromise on model capability or compliance, allowing you to build agentic applications for intelligence analysis or compliance automation. Evaluate these models on your specific workloads using the Bedrock console or SDKs, and implement scaling best practices for production readiness.

Key insights

AWS GovCloud (US) now offers open-weight FMs on Bedrock, ensuring secure, compliant AI for government missions.

Principles

• Data residency and zero operator access are paramount for regulated AI.
• Open-weight models enable independent security and behavior evaluation.
• Unified API access simplifies model selection for diverse use cases.

Method

Invoke models via "bedrock-mantle" (OpenAI SDK) or "bedrock-runtime" (AWS SDK), configure API keys/IAM, and implement retry logic for scaling.

In practice

• Use "bedrock-mantle" for OpenAI SDK compatibility.
• Adjust "reasoning_effort" for GPT OSS models.
• Implement exponential backoff for 503 errors.

Topics

• AWS GovCloud
• Amazon Bedrock
• OpenAI GPT OSS
• NVIDIA Nemotron
• Open-weight Foundation Models
• Government Compliance
• Serverless Inference

Code references

• openai/openai-python

Best for: AI Engineer, MLOps Engineer, AI Architect

Related on AIssential

• OpenAI models now available on Amazon Web Services — OpenAI's GPT-5.5, GPT-5.4, and Codex are now available on AWS Bedrock, integrating with existing AWS infrastructure and security.
• AI/BI Genie, Foundational Model API, and Databricks Assistant Now Generally Available in AWS GovCloud — Databricks now offers advanced generative AI and data intelligence features within AWS GovCloud for regulated U.S. government use.
• Why Nvidia builds open models with Bryan Catanzaro — NVIDIA's Nemotron project strategically advances AI infrastructure through open models, data, and research, driven by both internal product needs and ecosystem growth.
• AWS adds GPT-5.5, GPT-5.4 and Codex to Amazon Bedrock — AWS Bedrock now offers OpenAI's GPT-5.5, GPT-5.4, and Codex, providing enterprise access with integrated billing and unified APIs.
• OpenAI GPT-5.5 + Codex, now available and fully-governed in Databricks — Databricks integrates OpenAI's GPT-5.5 and Codex with Unity AI Gateway for governed enterprise AI applications.
• Run NVIDIA Nemotron 3 Nano as a fully managed serverless model on Amazon Bedrock — Nemotron 3 Nano offers an open, efficient, and accurate SLM for specialized agentic AI systems on Amazon Bedrock.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.