- Newest
- Most votes
- Most comments
Hello!
I send you the below rePost question previously answered for similar error regarding ThrottlingException, is for Claude v2.1 but it applies in the same way to Claude v3:
Here one of the possible solutions to this problem, provisioned throughput provides reserved capacity for your account specifically:
I'm also getting the ThrottlingException when trying to use Claude 3 Opus, regardless of when and how many times I try. Do I need to change some hidden configuration? Is this a known problem? Thank you.
Hi,
As you are using an on-demand model that makes use of a shared capacity pool, during periods of high demand across the service your account may be throttled below their expected rates.
The internal Bedrock team is working on expand capacity and address this issue, but currently we don't have an ETA in place for capacity increases.
Meanwhile you can try changing the model or the region if your use case allows it, or purchasing the provisioned throughput (see the upper link) to have allocated capacity to your account.
I'm having the same problem. Waiting for hours, and the agent is stuck throwing this error.
I have the same problem; every call I have tried to bedrock-runtime.invoke has failed with throttling. I have tried various regions and models; every single one fails. I'm trying via API and from a lambda function and with a simple example from the AWS documentation. What is the secret to get this to work ?
I manage two AWS accounts. One has premium support. More on that in a moment.
A problem is caused by AWS throttling their claimed default quota value for On-demand InvokeModel requests per minute for Anthropic Claude 3 Opus. Their published default quota value is 50. In both my accounts they throttled it down to 1 request per minute. This caused errors whenever a Claude 3 Opus agent triggered more than 1 Bedrock Builder Tools action group lambda function during a single interaction.
The problem was solved in about 3 days in my account with the premium support - they were slow, but they didn't quibble over setting my account to the default quota value of 50. Different story for my account without premium support. It's now day 5 of arguing with AWS billing support over their failure to supply the default quota value. They're asking ridiculous questions, wasting my time, being deliberately obstructive after I already justified my use cases. I've been using AWS since it was a beta service, before it even had a control panel. I'm disappointed in how AWS can no longer be trusted to be transparent and honest in their policies & pricing.
Relevant content
- asked 2 years ago
- asked 10 months ago
- asked 8 months ago
- asked 4 months ago
- AWS OFFICIALUpdated 7 months ago
- AWS OFFICIALUpdated 9 months ago
Hello. Specifically my question was when will this rate limit be removed or reset? As of now, it has been 18 hours and I am still receiving this rate limit/throttling error..
This is very difficult to work with, and still an issue despite "The internal Bedrock team is working on expand capacity and address this issue"
I'm barely using it - sometimes 2-3 requests within a minute will trigger this issue and fail. I'm not doing anything unusual / tricky - just calling these from Step Functions using the built-in Bedrock stages.
There must be a better option for Bedrock to scale up beyond 2-3 requests in a minute than having to purchase dedicated throughput? It's difficult to assess what the backoff algorithm should be, or which region to switch to, since there's almost no clarity on when this severe throttling will occur.
AWS must be able to scale better than this?