Indian teams evaluating Claude for a production app can now use Claude models through Amazon Bedrock with India geographic cross-Region inference. The key detail is that this is an AWS deployment route for developers and organizations—not a change to the consumer Claude.ai service.
AWS announced Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5 for this India inference profile on September 29, 2026. Here is what the launch means, how to try it, and what “in India” does—and does not—promise for your data.
What AWS announced
Amazon Bedrock lets developers call foundation models through AWS APIs and tools. With the new India geographic inference profile, an inference request can be handled in either of two AWS Regions: Mumbai (ap-south-1) or Hyderabad (ap-south-2). AWS says the profile keeps inference within India while drawing on capacity across both regions.
This can help teams that need to process prompts and responses within India for their own geographic or operational requirements. It is still important to check your organization’s legal, contractual, and compliance requirements separately; a regional cloud configuration is not, by itself, a blanket compliance certification.
How India cross-Region inference works
You choose a source Region for the Bedrock request. The India profile may route the request to Mumbai or Hyderabad, and prompts and generated output may move between those two Regions over AWS’s network. AWS states that customer data is not stored in the destination Region and remains in the source Region.
Bedrock’s default data security model is zero data retention for model inputs and outputs. AWS also notes that content flagged by automatic safety classifiers may be subject to human review for some models. Review the current Amazon Bedrock data protection documentation and your organization’s settings before sending sensitive data.
How to try Claude in the Bedrock console
- Sign in to the AWS Management Console and open Amazon Bedrock in Mumbai or Hyderabad.
- Check that your AWS account has access to the Claude model you intend to use.
- Open Playgrounds and choose the text playground.
- Select the India inference profile for the model, enter a non-sensitive test prompt, and run it.
- Compare the result, latency, and cost with your application’s requirements before integrating it.
The console playground is useful for an initial test without writing application code. For a production integration, use the Bedrock Runtime API, AWS SDK, or Anthropic Messages API support described in the AWS launch announcement. The exact model profile ID and regional availability can change, so copy the current value from the Bedrock console or AWS documentation rather than hard-coding an old example.
Before you build: a practical checklist
- Confirm access: Check model access and the India inference profile in the account and Region where your application will run.
- Use least-privilege IAM: Give the application only the permissions it needs to call the selected Bedrock model/profile.
- Test routing assumptions: The India profile can route between Mumbai and Hyderabad. Make sure this matches your own data-location requirements.
- Decide what to log: Avoid recording full prompts or outputs in application logs unless you have a clear retention and access policy. AWS CloudTrail and CloudWatch can help monitor activity and performance.
- Measure cost and latency: Test realistic prompts and expected traffic. Review the Amazon Bedrock pricing page because charges depend on the model and usage.
- Keep a fallback plan: Decide what your application should do if the selected model, profile, or Region is temporarily unavailable.
Does this mean Claude.ai is now hosted in India?
No. This announcement is about accessing Claude models through Amazon Bedrock for applications built on AWS. It does not announce a new Claude.ai consumer website or change the location settings of an existing Claude.ai account. If you are a developer, Bedrock gives you an AWS-managed API path; if you are an end user, the consumer Claude service is a separate product.
Who should consider this?
It may be relevant to Indian product teams, software companies, and enterprises already using AWS that want to build with Claude and keep inference within the India geographic profile. It may be less useful if you only need a personal chat interface, are not using AWS, or need a specific contractual guarantee that the profile does not provide. Start with a small, non-sensitive proof of concept and have your cloud or security team validate the data flow before production use.
Frequently asked questions
Which AWS Regions handle requests under the India profile?
AWS says requests route only between Mumbai (ap-south-1) and Hyderabad (ap-south-2). Prompts and output may move between those Regions.
Does AWS store my prompt in the destination Region?
AWS says customer data is not stored in a destination Region and remains in the source Region. Read AWS’s current data-protection documentation, especially if prompts can contain confidential or regulated information.
Can I use Claude without code?
Yes. AWS describes testing Claude in the Amazon Bedrock console text playground. For an application, use the supported Bedrock Runtime or Messages API flow.
Is Bedrock free?
Do not assume it is free. Pricing depends on the selected model and usage. Check current AWS pricing and set account budgets or usage alerts before running production traffic.
Bottom line
Claude on Amazon Bedrock’s India geographic inference profile gives AWS users a way to build applications with inference routed within Mumbai and Hyderabad. The practical benefits are access to shared regional capacity and an India-only inference path; the important caveat is that requests may move between those two Regions. Verify model access, data handling, IAM permissions, and pricing in your own AWS account before launch.
If you are comparing models for a different use case, see our guide to ChatGPT vs Gemini vs Claude for students.
