DeepSeek-R1 deployment o Azure AI foundry not responding, lags heavily and occassionally returns output.

Akinropo Taiwo 0 Reputation points
2025-02-11T00:43:43.2033333+00:00

I deployed a Deepseek-R1 model on Azure ai Foundry in the last week, initially, it was responding but really slow. Now it doesn't even respond anymore most times. It is super super slow, both via my application backend using streaming and azure llm playground as well.

What is wrong and how do i get this fixed? Is it a quota problem? or just a general capacity issue on azure ai Foundry?

Azure OpenAI Service
Azure OpenAI Service
An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
3,808 questions
0 comments No comments
{count} votes

1 answer

Sort by: Most helpful
  1. SriLakshmi C 3,255 Reputation points Microsoft External Staff
    2025-02-11T03:03:25.8066667+00:00

    Hello Akinropo Taiwo,

    Greetings and Welcome to Microsoft Q&A! Thanks for posting the question.

    I understand that you're experiencing a issue with the DeepSeek-R1 model on Azure AI Foundry, such as slow response times and occasional lack of output,

    I attempted to reproduce the scenario in my environment and encountered the same lagging issue. In some cases, the output was not generated at all.

    This could be due to quota limitations or capacity constraints. If your deployment has exceeded its assigned quota, it may result in throttling or significant delays in processing requests. Checking your quota limits and requesting an increase, if necessary, could help mitigate this issue.

    Additionally, capacity issues within Azure AI Foundry, such as high demand for resources or insufficient replicas and partitions, might be contributing to the performance degradation. If resource allocation is limited, it can lead to lags and unresponsiveness. To improve performance, consider scaling up your deployment by increasing the number of replicas and partitions.

    For further troubleshooting, check the Azure AI Foundry portal for any error messages, logs, or warnings that may indicate specific deployment issues.

    Please refer this https://github.com/nicholasdbrady/cookbook/blob/main/examples/deepseek/deepseek-r1-with-azure-ai-foundry-and-gradio.ipynb,

    I Hope this helps. Do let me know if you have any further queries.


    If this answers your query, do click Accept Answer and Yes for was this answer helpful.

    Thank you!


Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.