Confusing Model Behavior with Claude Vision and Q Developer CLI

Based on the information provided, there seem to be a couple of misunderstandings about the Claude models available in Amazon Bedrock and Amazon Q Developer CLI.

First, regarding the Claude models in Amazon Q Developer CLI, the system actually defaults to Claude 3.7 Sonnet when no specific model is selected. The model selection order in Q Developer CLI is:

1. Current session model selections (via /model or --model)
2. User-configured preferences in settings
3. System default (Claude 3.7 Sonnet)

The screenshot you shared might be showing an outdated version of the CLI or there might be a configuration issue. The latest version of Amazon Q Developer CLI (v1.11.0) supports Claude Sonnet 4, which you can select during active chats using the /model command or set as your default model using q settings chat.defaultModel claude-4-sonnet.

Regarding the OCR issues with Claude Vision in Lambda, there are differences in how Claude 3.7 Sonnet and Claude 4 models handle thinking and output. Claude 4 models return summarized thinking rather than full thinking output, and they handle redaction differently. This might explain some of the unexpected behavior you're seeing.

To troubleshoot your Q Developer CLI issue, I'd recommend:

1. Checking your CLI version and updating if needed
2. Verifying your AWS region supports the Claude 4 model
3. Using the /model command to explicitly select Claude Sonnet 4

For the OCR issues in Lambda, you might want to check if you're properly configuring the model parameters and ensure you're using the appropriate API calls for the specific Claude model version you're working with.
Sources
Differences in thinking across model versions - Amazon Bedrock
Access Claude Sonnet 4 in Amazon Q Developer CLI | AWS DevOps & Developer Productivity Blog
Anthropic's Claude : analyse graph images and answer questions about them | AWS re:Post

Update on Claude Vision Model Confusion

I’ve traced the source of the confusion with Claude Vision back to the actual model itself—it appears to be working as expected. The only remaining issue is that the Q Developer CLI still does not recognize or use the 4.0 model and instead defaults to version 3.5.

Good afternoon,

It may be a limitation of the Region you are currently attempting to use in regards to Claude 4 availability.

Some things to check:

Check the CLI version and update if needed(Developer CLI (v1.11.0) is latest)
Verify if the AWS region supports the Claude 4 model
Use the /model command to explicitly select Claude Sonnet 4

Hey Matt, All things that I've done. I went back to check region and began to think about how I sometimes sign-in with my Builder ID and sometimes with Management Console. My Builder ID is set with us-east-1, but my Management Console default region is set to last used in my unified settings.

I won't delete my JWT token tonight to check but I've reset my default to us-west-2 and we'll see if it's working tomorrow. Thanks for the head check. It would be great to check Q Developer cli region from Q Developer. I'll start a feature request. All the best.

I used my Builder ID that is linked to us-east-1 to sign-in, Q Developer is still reporting to use sonnet-4 but the model is reporting that it's version 3.5.