How to Fix AWS Account Stuck at 0 TPM/RPM for Anthropic Models in EU-Central-1

Select Language:

If you’ve just created a new AWS account and are running into issues with Bedrock quotas being set to zero, you’re not alone. Many users report that their account shows no usage limits (0 requests per minute and 0 tokens per minute), which prevents them from making any requests to models like Anthropic Claude.

Here’s a step-by-step guide to help fix this problem and get your quotas set correctly.

First, double-check that your account has proper access. Ensure that your IAM permissions include the necessary rights to invoke the Bedrock models, like bedrock:InvokeModel, bedrock:InvokeModelWithResponseStream, bedrock:Converse, and bedrock:ConverseStream. Also, confirm that the models appear as enabled in your console. You can do this by using the IAM policy simulator or the console’s model list.

Next, it’s important to verify that your account has been seeded with the default quotas. When a new account is created, sometimes the initial quotas are not automatically assigned, which causes the zero-limit issue. This means every request you try to make results in a throttling error, such as “Too many tokens per day, please wait,” regardless of your request size.

If you’ve already tried requesting a quota increase by opening an AWS Support case, but it still hasn’t been assigned or responded to after more than a day, don’t worry. This can happen with new accounts, especially if no usage has been recorded yet.

To move forward, consider these options:

Reach out to AWS Support again: Emphasize that your account is new, and you need baseline quotas to start using Bedrock. Since your account doesn’t have any usage history, request that AWS pre-seed your account with the default quotas to enable you to begin testing and development without delays.
Provide detailed use case justifications: When communicating with support, explain that your account will be used to serve GDPR-compliant inferences for a European client, with strict data residency requirements. Clearly stating this can help prioritize your request.
Check for documentation or community advice: Though there may not be a formal process published, some users have successfully escalated these issues by explaining the critical need for initial quotas. Posting on forums like re:Post or engaging with AWS support representatives directly may lead to faster assistance.
Understand the importance of seeded defaults: In some cases, AWS can seed new accounts with baseline default quotas upon request. If this option is available, support can set these quotas without requiring a standard quota increase, saving time and allowing you to start testing immediately.

Dealing with quota issues on a new account can be frustrating, especially for essential projects like GDPR-compliant AI inference. Persistence is key. Keep communicating clearly with AWS Support, providing your use case details, and emphasizing the urgency based on your project timeline.

This approach has helped others overcome similar issues, and with some persistence, you should be able to get your account properly configured to start using Bedrock models effectively.