Share via

Grok-4-fast-reasoning Serverless (MaaS) deployment timing out in Korea Central

벨루가 0 Reputation points
2025-12-12T08:35:46.1333333+00:00

Hello,

I am using Azure AI Foundry with an Azure Direct Model deployment of Grok-4-fast-reasoning.

  • Subscription: ISV 벨루가 서비스 구독
  • Resource name: pnu-resource (Microsoft Foundry)
  • Region: Korea Central
  • Project name: proj-pnu-grok
  • Deployment names:   - grok-4-fast-reasoning

Symptoms:

  • In the Azure AI Foundry playground, chat requests to these deployments keep running for a long time and eventually time out.
  • When calling the same deployments via the Python SDK (chat.completions.create with the generated endpoint and key), the requests also time out.
  • In Monitoring/Observability, I can see requests logged with input tokens counted, but almost no output tokens (for example, ~1,000 input tokens vs. 0–6 output tokens across about 15 requests).
  • The deployment provisioning state is Succeeded.
  • I deleted and recreated the Grok-4-fast-reasoning deployment, but the behavior is the same.
  • Other models in the same resource/region (for example, Grok-4-fast-non-reasoning) do not show this issue.

It looks similar to other reports of Grok-4-fast-reasoning timeouts in Azure AI Foundry.   Is there a known issue with Grok-4-fast-reasoning in Korea Central, or any recommended workaround (for example, a different region) for production use?

Azure AI services
Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
0 comments

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Q&A Assist
    2025-12-12T08:35:56.7466667+00:00
    AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more

    It seems you are experiencing timeouts with the Grok-4-fast-reasoning model deployed in the Korea Central region. While the provisioning state indicates success, the symptoms you describe, such as long-running requests and minimal output tokens, suggest there may be an issue specific to this model in that region.

    Currently, there is no specific mention of known issues with Grok-4-fast-reasoning in Korea Central in the provided context. However, if other models are functioning properly in the same resource and region, it may be worth considering deploying the Grok-4-fast-reasoning model in a different region where it has been reported to work without issues. This could potentially resolve the timeout problems you are encountering.

    For further assistance, you might want to check Azure's support channels or documentation for any updates regarding this model's performance in specific regions.

    References:

    0 comments
  2. 벨루가 0 Reputation points
    2025-12-12T08:41:19.48+00:00

    Thanks for the suggestion. According to the region availability matrix, Grok-4 is available in Korea Central and other regions.

    I will try deploying Grok-4-fast-reasoning in another supported region (for example, East US 2 or Sweden Central) and see if the timeouts still happen. However, since the issue reproduces consistently in Korea Central (Playground + SDK) while other models work fine, it would be helpful if the product team could also check the runtime for Grok-4-fast-reasoning in Korea Central for my resource (pnu-resource, proj proj-pnu-grok).

Your answer

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