Amazon Bedrock

Amazon Bedrock is AWS's managed platform for building generative AI applications and agents at production scale. AWS describes it as giving access to hundreds of foundation models from leading AI companies, plus evaluation tools to pick a model on performance and cost. It is where teams already on AWS reach models such as Anthropic's Claude and OpenAI's GPT-6.1 Sol without leaving their cloud account.

Key features

• Model choice: one API surface for models from several providers. On September 29, 2026 AWS made GPT-6.1 Sol generally available there, and extended Claude Opus 5 and Sonnet 5 inference to Seoul and Singapore.
• In-region and in-country inference: for Claude in Seoul and Singapore, AWS says the request is processed entirely inside the Region you call. In India, geographic cross-Region inference routes only between ap-south-1 and ap-south-2.
• Guardrails: AWS documents Bedrock Guardrails for filtering harmful content and checking responses, and they work with the in-region Claude models.
• Customization: Knowledge Bases, Bedrock Data Automation, prompt engineering and fine-tuning for using your own data.
• Agents: Amazon Bedrock AgentCore is AWS's platform to build, connect and optimize agents with any framework and model.
• Security: AWS states data is encrypted in transit and at rest and that Bedrock does not store or use your data to train models.

Use cases

• Enterprises with data-residency rules that need Claude inside a specific Region.
• Teams that want several model vendors behind one AWS bill and IAM policy.
• Building RAG assistants and agents with Knowledge Bases and AgentCore.

Pricing

Pricing depends on the model and mode. For the in-region Claude models AWS says billing follows standard on-demand pricing for the Region you call, with per-Region service quotas. See the official Bedrock pricing page for current numbers.

Quick start

1. Enable model access in the Bedrock console for your Region.
2. Call the model through the Converse, InvokeModel or Messages API on the bedrock-runtime endpoint.
3. Add a Guardrail and test with your own prompts before production.
4. Check the quota for the Region you call, since quotas are per Region.

Limitations

Model availability differs by Region and by inference type. Cross-Region inference and single-Region inference give different guarantees, so read the model page for your Region before promising residency to customers.

FAQ

Is Bedrock the same as calling the model vendor directly?

No. You call AWS's endpoint, so billing, quotas, IAM and Region rules are AWS's, and feature timing can differ from the vendor's own API.

Which Claude models are in Seoul and Singapore?

Per AWS, Claude Opus 5 and Sonnet 5 in Seoul, and Sonnet 5 in Singapore.

Alternatives

• OpenAI: use the model vendor's own platform.
• GitHub Copilot: a product surface that fronts many of the same models.

Conclusion

If your workloads and data already live on AWS, Bedrock is the shortest path to production-grade Claude and GPT models with regional controls. Start from the official Bedrock page.