How can I resolve issues when I encounter a "model not found" or "unsupported model" error in Bedrock?

When encountering a “model not found” or “unsupported model” error in AWS Bedrock, the issue typically stems from incorrect model identifiers, regional availability constraints, or missing access permissions. Start by verifying the exact model ID you’re using. Bedrock requires precise model names, such as anthropic.claude-v2 for Claude 2 or amazon.titan-text-lite-v1 for Amazon Titan. Typos, incorrect version numbers, or using a model alias that isn’t supported (e.g., claude-v2 instead of the full ID) will trigger errors. Check the AWS documentation for the latest valid model IDs and ensure your code or API calls match them exactly. Also, confirm that the model is available in your AWS region—for example, Claude 2 isn’t supported in all regions, so using us-east-1 might work while another region could fail.

Next, ensure you’ve been granted access to the model in the Bedrock console. AWS requires explicit enablement for many models, even if they’re available in your region. Navigate to the Bedrock dashboard, select “Model access” in the left menu, and check if the model is listed as “Access requested” or “Not requested.” If access hasn’t been granted, submit a request via the console. For example, accessing the Jurassic-2 model from AI21 Labs requires manual approval. Additionally, verify that your IAM policies grant the bedrock:InvokeModel permission for the specific model ARN. A misconfigured policy that omits the model’s resource ARN (e.g., arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-v2) will block access even if the model is otherwise available.

Finally, consider API version mismatches or deprecated models. If you recently updated your SDK or CLI tools, ensure your code aligns with the latest API requirements. For example, Bedrock’s InvokeModel API expects specific parameters like accept and contentType headers, and some models require JSON-formatted input (e.g., Anthropic Claude models need a prompt field in the body). If the model was working previously but suddenly fails, check the AWS Service Health Dashboard or Bedrock documentation for deprecation notices. For instance, older Titan model versions might be phased out in favor of newer iterations like titan-text-express-v1. If all else fails, use the AWS CLI to list available models in your region with aws bedrock list-foundation-models --region us-east-1 --query "modelSummaries[*].modelId" to validate availability and identifiers.