- Newest
- Most votes
- Most comments
It sounds like you may be hitting anomaly prevention.
Quotas protect customers and AWS. For example, a Bedrock TPM limit prevents customers from accidentally running up a huge bill. Additionally, AWS
Is this a new account that was recently setup?
Hi there,
Sorry to hear about this.
We've reviewed your support cases and left internal notes expressing your urgency. We're unable to discuss any account-specific info over social media, so keep an eye on your cases for updates. The cases have been assigned and the support agent will respond as soon as possible.
- Reece W.
I understand your frustration with the quota provisioning issue. While I can't directly escalate your support cases, I can provide some context that might help you troubleshoot and communicate with AWS Support more effectively.
The throttling you're experiencing with the "Too many tokens per day" error, even with a simple "Say Hello" prompt, is likely related to how Amazon Bedrock reserves quota at the start of each request. When you make a request without explicitly setting the max_tokens parameter, Bedrock defaults to the model's maximum output capacity. For Claude models like Sonnet, this can be as high as 64,000 tokens.
This means that even though your actual prompt and expected response are tiny, the system reserves quota based on the maximum possible output. For example, a request with 1,000 input tokens and no max_tokens set would reserve 65,000 tokens (1,000 input + 64,000 default max output) at the request start, even if the actual output is only 100 tokens.
However, your core issue appears to be that your account quotas are showing 0 TPM/0 RPM instead of the published defaults. This is indeed a provisioning issue that requires AWS Support intervention. For your support cases:
- Emphasize that this is a quota provisioning issue, not a usage issue
- Specify that you need the account-level quotas initialized to AWS default values
- Reference the specific Service Quotas you're seeing (0 vs. the 5,000,000 TPM default)
- Consider opening a new case specifically categorized under "Account and Billing Support" > "Account" rather than technical support
Since you're on Basic Support and experiencing delays, you might also try reaching out through the AWS re:Post community or consider temporarily upgrading to Developer Support for faster response times on account-level issues.
Sources
TPM & RPM Quota Monitoring Dashboard for Amazon Bedrock | AWS re:Post
The technical explanation regarding token reservation is accurate and helpful for overall understanding. However, it doesn't directly address the initial provisioning failure where the quota is stuck at 0 instead of the 5M default. The advice on support routing is practical and appreciated
Relevant content
- AWS OFFICIALUpdated 8 months ago
All my 5 cases were closed with the same comment :
Dear Customer,
We've reviewed your Bedrock quota increase request. At this time, we recommend building more account history before we can process this type of increase. Our quota management system helps ensure a smooth scaling experience while protecting your account from unexpected costs. To better evaluate future requests, we recommend:
Utilising your existing AWS services more extensively Establishing a consistent usage pattern over at least one billing cycle(s) Once you have built a more comprehensive usage history, we welcome you to submit a new request. For enterprise-scale implementations, consider:
This is not really correct, as my quotas were set to 0 from the provisioning time. So Unexpected cost would never be an issue. The way it was provisioned is an issue here.
I reopened those cases 21 days ago with confirmation that there was a bill payment established on this account, so please proceed with fixing my quotas to the default ones as they should be. Just to make it clear, I am not asking for a quota increase just to be provided with the default ones as per AWS