- Newest
- Most votes
- Most comments
This looks like a quota issue on the surface, but based on the behaviour you described, it’s unlikely to be real quota exhaustion.
If you’re getting HTTP 429 “Too many tokens per day” on first or very low usage, especially across multiple regions, it’s usually a provisioning issue with Amazon Bedrock rather than actually hitting a limit.
I’ve seen this happen on newly enabled accounts where the effective quotas for Anthropic Claude models weren’t properly initialised, so the system behaves as if your token limit is already exhausted. To help AWS Support resolve it faster, include the request IDs (x-amzn-requestid) from failed calls so they can trace the issue internally. It’s also worth asking them to verify your effective quotas (not just defaults) for Claude models in both eu-west-2 and us-east-1, specifically tokens per day and on-demand tokens per minute.
You should also check the Service Quotas console yourself. If anything shows as 0 or unusually low, attach a screenshot to your case. Even if the quotas appear correct, submitting a quota increase request can sometimes trigger backend provisioning if it was missed during setup.
The key signal here is that it’s happening on first or very low usage, across multiple regions and multiple Claude variants. That combination strongly points to an account-level configuration issue rather than normal throttling.
Once Support checks the backend quota state, they should be able to correct it without further changes on your side.
Based on the 429 "Too many tokens per day" error you're experiencing with Amazon Bedrock Anthropic Claude models, this is indeed a quota-related issue that requires attention. Here's what you need to know:
Amazon Bedrock has service quotas that control your usage, and these quotas vary by model and region. The error you're seeing indicates you've hit the tokens-per-day limit for the specific Claude model you're trying to use. These quotas are adjustable, but they require a formal request through AWS Support.
To resolve this issue and speed up your support case, you should:
Collect the following information:
- The specific model IDs you're trying to use (e.g., anthropic.claude-v2, anthropic.claude-3-sonnet, etc.)
- The AWS regions where you're experiencing the issue
- Request IDs from failed API calls (found in the response headers)
- Your expected usage patterns and token requirements
Request a quota increase: You can request quota increases through the Service Quotas console or directly through your AWS Support case. Navigate to Service Quotas, search for "Bedrock," and look for the specific quota limits related to the Claude models you need. Each model has its own quota for tokens per minute and other rate limits.
Important considerations:
- Default quotas for Bedrock models can be quite restrictive initially
- Quota increases are evaluated based on your use case and account history
- Different Claude model variants (Haiku, Sonnet, Opus) have separate quotas
- Processing times for quota increase requests can vary
Since you've already opened a support case, make sure to provide them with the specific model ARNs, regions, and your business justification for higher limits. This will help expedite the review process.
Hello.
Is your AWS account recently created?
I often see questions on AWS re:Post from people with similar problems to yours.
For new AWS accounts, Bedrock quotas are often set to a very low number (close to zero).
The response at the following URL suggests that requests may be rejected depending on factors such as AWS account usage.
Since you've already contacted AWS support, I think you'll just have to wait for their response.
https://repost.aws/ja/questions/QUf16LkLwNS2yRbKVufdu6FA/new-account-stuck-at-0-tpm-rpm-for-all-bedrock-models-despite-aws-default-showing-5-000-000-tpm-provisioning-issue
https://repost.aws/ja/questions/QUfFix7mDkRIKGpgLRxYMGaA/throttlingexception-too-many-tokens-per-day-please-wait-before-trying-again
Relevant content
- AWS OFFICIALUpdated a year ago
