How can we test the Australia East replacement for GPT‑4.1‑mini before retirement?

Fadi Sarkis 0 Reputation points
2026-11-10T21:39:29.17+00:00

We run GPT‑4.1‑mini on a Standard deployment in Australia East. Our retirement notice gives 14 April 2027 as the end date. We need to keep production processing in Australia and test a replacement separately for invoice images, structured output and tool calling.

Has Microsoft published the replacement model and version for existing Australia East Standard deployments? We’re looking for a generally available model with at least 12 months of support from migration. Foundry currently shows 0/0 TPM for GPT‑5.6‑Sol Standard and 0/0 TPM for all Claude models too...

If the successor hasn’t been announced, where will Microsoft publish it, and how can an existing customer obtain a small test deployment before the automatic upgrade?

Foundry Models
Foundry Models

A catalog of AI models in Microsoft Foundry that you can discover, compare, and deploy using Azure’s built‑in tools for evaluation, fine‑tuning, and inference

0 comments

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Fadi Sarkis 0 Reputation points
    2026-11-10T21:48:18.31+00:00

    Could a Microsoft moderator clarify which upgrade applies to our existing Australia East Standard deployment? Our Azure retirement email names GPT‑5‑mini (2025‑08‑07) as the regional replacement for GPT‑4.1‑mini, but the public retirement schedule leaves the replacement field blank.

    Our subscription reports Free Tier for model quota, and Foundry shows 0/0 TPM for GPT‑5.6‑Sol Standard in Australia East. Given the published restriction on Free Tier quota increases, is there a supported way for an existing customer to test the regional successor in a separate deployment before the automatic upgrade?

    We need production processing to stay in Australia and a model with at least 12 months of published support from migration. If Microsoft hasn’t announced an option meeting both requirements, could you confirm that and tell us where the regional migration path will be published?

    Was this answer helpful?

    0 comments

  2. AI answer

    2026-11-10T21:41:12.3966667+00:00

    Microsoft has not yet named the official GPT‑4.1‑mini replacement; request a small Australia East Standard quota allocation now and test GPT‑5.6‑Sol side by side, but do not treat it as the successor or as meeting a 12‑month migration runway unless migration finishes by 11 January 2027.

    Current published position

    • gpt-4.1-mini version 2025-04-14 retires on 14 April 2027.
    • Its Replacement field is currently blank. Microsoft publishes the selected replacement in the Model retirement schedule, normally about 90–120 days before retirement. That schedule also identifies the automatic-upgrade target where supported.^1^
    • Your existing Standard deployment is eligible for a Microsoft-managed, rolling regional upgrade at retirement. The replacement can be made available through that upgrade even if customers cannot deploy it separately in that region beforehand.^2^
    • Do not rely on automatic upgrade as the testing plan: model changes can affect JSON shape, image extraction, latency, tool calls, prompts, and downstream code.^3^

    Candidate available for regional testing

    The current regional matrix lists GPT‑5.6‑Sol 2026-07-09 for Standard/Regional in Australia East. Standard/Regional means inference is processed in the deployment’s region; this differs from Global Standard and APAC Data Zone processing.^4^

    GPT‑5.6‑Sol supports image input, structured outputs, functions/tools, and parallel tool calling, so it is technically suitable for exercising the stated invoice workflow.^5^

    However:

    • It is a candidate, not the published GPT‑4.1‑mini replacement.
    • Its published retirement date is 11 January 2028.^1^
    • Therefore, it provides at least 12 months after migration only if production migration completes on or before 11 January 2027. Migrating near 14 April 2027 would leave only about nine months. If that deadline is impractical, wait for a later-retiring GA candidate or obtain written lifecycle guidance from Microsoft before selecting the production target.

    Obtain the test deployment

    Because Foundry shows 0/0 TPM, model availability does not currently give the subscription deployable quota.

    1. In the new Foundry portal, select the correct subscription and project.
    2. Go to Manage → Quota → Token per minute.
    3. Filter for:
      • Region: Australia East
      • Model: GPT‑5.6‑Sol
      • Deployment type: Standard (documentation may call this Standard/Regional)
    4. Select Request quota and request the smallest practical TPM allocation. Viewing quota requires Cognitive Services Usages Reader; requesting it requires Owner or Contributor at subscription scope.^6^
    5. After allocation, create a separate deployment through Build → Catalog → Model Catalog → GPT‑5.6‑Sol → Deploy → Custom settings. Confirm before deployment that its details explicitly show:
      • Standard
      • Australia East
      • model version 2026-07-09
    6. Keep GPT‑4.1‑mini serving production. Replay a frozen invoice test set against both deployments and compare image extraction, schema validity, tool-selection accuracy, tool arguments, latency, and cost.^3^

    Quota approval and physical Australia East capacity are separate. A request can be submitted, but approval does not guarantee immediate regional capacity. If the quota remains 0/0, or allocation is approved but deployment still fails, open an Azure Support request and include the subscription, region, model/version, requested TPM, deployment type, timestamp, and correlation/error ID.

    Foundry shared quota may be used for temporary testing, but only use it if the resulting deployment explicitly identifies Standard/Regional and Australia East. Otherwise, its processing boundary is not established as Australia East and it does not meet the onshore requirement.^6^

    Continue monitoring the retirement schedule and the Models API fields lifecycleStatus, deprecation, and per-SKU deprecationDate; the schedule is where Microsoft will publish the official replacement and upgrade target.^2^

    References

    1. Model retirement schedule
    2. Microsoft Foundry Models lifecycle and support policy
    3. Model migration process
    4. Region availability for Foundry Models sold by Azure (standard)
    5. Azure OpenAI reasoning models
    6. Manage and increase quotas for resources with Microsoft Foundry (Foundry projects)
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

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.