Skip to content
Nubent

/Glossary · AI

What is Claude?

Claude is the family of large language models developed by Anthropic. It can reason over text, code, and images, and it can use tools. On AWS, you access it through Amazon Bedrock, inside your own account and under your IAM controls.

How Claude works and which models it includes

Claude comes in several model lines, such as Opus, Sonnet, and Haiku, each with a different balance of reasoning capability, speed, and cost per token. All of them accept system prompts, multi-turn conversations, image inputs, and tool definitions the model can call to look up data or take actions.

Anthropic ships new versions often, so pin the exact model ID in your configuration and test every version change against your own evaluation set before it reaches production. Picking the smallest model that handles the task well usually cuts both latency and cost.

Claude on AWS through Amazon Bedrock

Amazon Bedrock exposes Claude models as a managed service: you call the model through Bedrock APIs such as InvokeModel or Converse, authenticated with IAM roles and with no servers to manage. The latest models are served through cross-region inference profiles, which spread load and improve availability.

Running inside your AWS account simplifies security and governance: traffic can go through VPC endpoints with PrivateLink, usage is logged in CloudTrail, metrics flow to CloudWatch, and spend shows up on your bill next to every other service, with tags to allocate it by team.

What Claude is used for in products and engineering

In products, Claude powers assistants over internal documentation with RAG, data extraction from contracts or invoices, ticket classification, and agents that run multi-step workflows with tools. What matters most is defining the context it receives, the tools it can use, and how its output is validated, so untrusted input cannot steer it into leaking data.

In engineering, Claude writes and reviews Terraform, proposes pipeline changes, summarizes incidents from logs, and drafts technical documentation. At Nubent, every agent output goes through a pull request and is approved by an engineer before it reaches production.

/FAQ

Questions about Claude

Another question? Write to hello@nubent.com and an engineer will answer.

How do I use Claude on AWS?

Enable access to Anthropic models in the Amazon Bedrock console, grant invoke permissions with an IAM policy, and call the model with the AWS SDK through the InvokeModel or Converse APIs. For recent models, you use an inference profile ID.

What is the difference between Claude on Bedrock and the Anthropic API?

The model comes from Anthropic either way. On Bedrock, authentication, networking, audit logs, and billing stay in your AWS account, which makes internal policies easier to meet. The Anthropic API uses its own keys and may get some features first.

Which Claude model should I choose?

It depends on the task. Larger models do better at complex reasoning and code, while smaller ones respond faster and cost less per token. The practical approach is to build an evaluation set from real cases and compare quality, latency, and cost before deciding.

Does Claude train on the data I send through Bedrock?

According to AWS documentation, Amazon Bedrock does not use your prompts or responses to train models and does not share them with model providers. You should still review the current terms and apply encryption, access controls, and usage logging.

Next step

What should we build next on AWS?

30 minutes with an engineer. Concrete next steps, whether or not you work with us.