12
Tell HN: Claude Opus 4.7 quota suddenly changed to 0 TPM in Bedrock
Suddenly our Opus 4.7 access was removed from Bedrock ( The quota was set to 0 suddenly).
This isn’t the first time I’ve faced this issue. Is anyone else experiencing the same problem?
Looks like AWS can revoke access to frontier models anytime without any warnings. The lack of transparency is not the right thing to do. The quality of AWS support and services used to exceptional.
Here’s an update we received from AWS support.
“ Following a thorough investigation with our internal teams, we have identified what occurred with your Claude Opus 4.7 model access. Your account previously had access to this model and was successfully using it until yesterday. However, model access for a given account can be updated based on different factors including regional considerations, payment history, and usage patterns to maintain the performance of the service and ensure appropriate usage of Amazon Bedrock.
A recent system update adjusted access controls, which resulted in your account's quota being set to 0 starting May 1st, 2026. This is why your applications began encountering throttling errors today.
As part of this escalation, I have reached out to your AWS Account Manager to explore options for access restoration. However, I want to set proper expectations that we cannot guarantee approval for access restoration requests, as accessibility is subject to change automatically based on various factors.
To unblock your production impairment while we pursue the access restoration review, I strongly recommend migrating to an alternative model such as Claude Opus 4.6 [1]. Your current quotas for this model in us-east-1 are:
Cross-region model inference requests per minute for Anthropic Claude Opus 4.6 V1: 10,000 Cross-region model inference tokens per minute for Anthropic Claude Opus 4.6 V1: 3,000,000 Global cross-region model inference requests per minute for Anthropic Claude Opus 4.6 V1: 10,000 Global cross-region model inference tokens per minute for Anthropic Claude Opus 4.6 V1: 3,000,000 The Opus 4.6 model can serve as an effective replacement with minimal code changes, allowing you to restore service to your government customers immediately.”
I was noticing token counts on opus 4.7 which seemed at least 2x what they should have been. I wonder if they are doing a fix ?
Insane for a company to pull this crap on paying customers with production workflows.
Amazon bedrock randomly throttles you. Quinnypig has talked about this extensively and how many times they've rugpulled even those with enterprise support for their critical production systems.
It's a bad inference provider. Consider moving to Google or Claude itself.
Any alternative to bedrock?
I'm ok to pay for GPUs and currently trying vast.ai, I just need the firepower and I'll use it for my org's usage and the way I want to using opensource models.
I don't want to policing or get locked down or throttled based on my usage or volume.
One flat fee and pricing.
The reason is that I currently don't have the necessary infra available (though working towards acquiring GPUs), I want the dev to continue without any bottlenecks.
AWS GPU offerings are distinct from Bedrock. This is specifically about AWS Bedrock and using an inference provider that gives access to Claude.
If you just want GPUs, just talk capacity with an account manager at your favorite hyperscaler.
They quietly changed Opus 4.7 quota to 0. I can lo longer make even a single request