AWS Bedrock Remove Service Llama 3.2 1B Instruct

The error you're encountering when trying to remove access to the Llama 3.2 1B Instruct model in AWS Bedrock is expected behavior. Certain models in Amazon Bedrock, including Meta's Llama models, cannot be disabled once they have been enabled. This is a deliberate design choice by AWS.

Specifically, you cannot remove access from Amazon Titan models, Mistral AI models, or the Meta Llama 3 Instruct models once they have been activated. This includes the Llama 3.2 1B Instruct model you're trying to remove.

While you can't directly remove access to these models through the Bedrock console, there is an alternative approach if you need to restrict usage. You can use an identity-based IAM policy to deny access for running inference on specific models. This allows you to effectively prevent users from using the model without actually removing it from your account.

If you're concerned about costs, it's worth noting that even though you can't remove access, you are only charged for actual usage of the model (e.g., when you make inference calls), not for simply having access enabled.

If you need to manage which models appear in your list of available models for programmatic access, you may need to implement additional logic in your code to filter out the models you don't want to use, even if they remain enabled in your Bedrock account.
Sources
Bedrock Model access modifications | AWS re:Post
How to remove access for Bedrock Models? | AWS re:Post