Share via

Inquiry regarding the impact of Azure AI Inference SDK (preview) deprecation

85795786 160 Reputation points
2026-03-30T01:47:42.4+00:00

Note: This message has been translated from Japanese to English using a translation tool. Please excuse any unnatural phrasing or terminology. Thank you for your understanding.

I am writing to clarify the impact of the upcoming deprecation of the Azure AI Inference Beta SDK on our current production environment.

According to the "Azure AI Foundry model inference endpoints" documentation, the Azure AI Inference Beta SDK is scheduled for retirement on May 30, 2025.

Our organization currently utilizes Azure OpenAI Service in the environment described below. We would appreciate it if you could clarify whether our setup falls within the scope of this deprecation or if we can continue using it as-is, specifically regarding the following three points:

1. Usage of the Python SDK (openai package)

We have implemented our solution using the openai library as follows:

Python

from openai import AsyncAzureOpenAI

Our understanding is that the deprecation specifically targets the beta version of the azure-ai-inference package. Since we are using the openai package (Azure OpenAI SDK), could you confirm that our implementation will not be affected by this deprecation?

2. Usage of API Version 2024-10-21

We are currently using API version 2024-10-21. While the documentation recommends migrating to the v1 API, do we need to complete any API version changes or migration tasks by the May 30, 2025 deadline? Furthermore, should we treat the SDK deprecation and the API version lifecycle as independent schedules?

3. Impact on Azure OpenAI Resources and Endpoints

Currently, some of our resources are created as traditional "Azure OpenAI Service" resources rather than through Microsoft Foundry.

Endpoint Format: https://[resource-name].openai.azure.com/

In light of these changes, is there any possibility that we will be required to switch to Microsoft Foundry, change our endpoint URLs, or recreate our resources?

The original Japanese text follows below:

Azure AI Inference SDK (preview) の廃止予定に関連し、現在運用中の環境への影響について確認させてください。

ドキュメント「Azure AI Foundry モデル推論エンドポイント」にて、Azure AI 推論ベータ SDK が2025年5月30日に廃止される旨の案内を拝見いたしました。

現在、弊社では以下の環境で Azure OpenAI Service を利用しておりますが、

今回の影響範囲に含まれるのか、あるいは継続利用が可能か、以下の3点についてご教示いただけますでしょうか。

  1. Python SDK（openai パッケージ）の使用について

現在、以下の通り openai ライブラリを使用して実装しております。

from openai import AsyncAzureOpenAI

今回の廃止対象は「azure-ai-inference パッケージのベータ版」という認識ですが、

上記のように openai パッケージ（Azure OpenAI SDK）を利用している場合は、

今回の廃止による影響はないという認識で相違ないでしょうか。

  1. API バージョン 2024-10-21 の利用について

現在、APIバージョンに 2024-10-21 を指定して利用しております。

ドキュメントでは v1 API への移行が推奨されていますが、今回の SDK 廃止期限である 2025年5月30日までに、

APIバージョンの変更や移行作業を完了させる必要はありますでしょうか。 

もしくは、SDK の廃止と API バージョンのライフサイクルは別という扱いになりますでしょうか。

  1. Azure OpenAI リソースおよびエンドポイントへの影響について

現在、Microsoft Foundryとしてではなく、従来の「Azure OpenAI Service」リソースとして作成したものを一部使用しております。

エンドポイント形式：https://[リソース名].openai.azure.com/

今回の変更に伴い、Microsoft Foundry への切り替えや、エンドポイント URL の変更、

あるいはリソースの再作成などの対応が必要になる可能性はございますでしょうか。

Azure OpenAI in Foundry Models
0 comments

Answer accepted by question author

  1. SRILAKSHMI C 18,225 Reputation points Microsoft External Staff Moderator
    2026-03-30T03:17:00.26+00:00

    Hello 85795786,

    Welcome to Microsoft Q&A and Thank you for your detailed questions, I understand you want to ensure there is no impact to your current production environment due to the Azure AI Inference SDK (preview) deprecation.

    Based on your setup, there is no immediate impact, and your current implementation can continue to run as-is. Please find the breakdown below:

    1. Usage of Python SDK (openai package)

    Your understanding is correct.

    You are currently using:

    from openai import AsyncAzureOpenAI

    The deprecation announced for May 30, 2025 applies specifically to the Azure AI Inference preview SDK (azure-ai-inference beta)

    The official OpenAI SDK (openai PyPI package) is not part of this deprecation

    This SDK communicates using the /openai/v1/ API under the hood

    Your current Python implementation using the openai package will continue to function normally. No changes are required.

    2. API version (2024-10-21) and v1 API migration

    In your case:

    • When using the OpenAI SDK, the API version is abstracted and handled internally
    • The SDK already uses the /openai/v1 API, so you do not need to manually manage api-version
    • The SDK deprecation (preview package) and API version lifecycle are independent

    There is no requirement to migrate API versions by May 30, 2025

    You can plan any API upgrades separately based on your roadmap

    3. Azure OpenAI resources and endpoints

    Your current setup:

    https://<resource-name>.openai.azure.com/

    is part of the classic Azure OpenAI Service resource type, and:

    • These resources are not being deprecated
    • The Azure AI Foundry model inference endpoints are a separate offering
    • There is no mandatory migration to Foundry

    You are not required to Change endpoint URLs, Recreate resources, Move to Foundry

    You can continue using your existing Azure OpenAI resources and endpoints without any changes.

    No impact from Azure AI Inference SDK (preview) retirement

    No changes required for your current SDK (openai) usage

    No forced API version migration

    No changes required for your Azure OpenAI resources or endpoints.

    Please refer this

    Migrate from Azure AI Inference SDK to OpenAI SDK (Python) https://learn.microsoft.com/azure/foundry/how-to/model-inference-to-openai-migration?tabs=python-key

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

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

    Thank you!

    Was this answer helpful?

    1 person found this answer helpful.
    0 comments

1 additional answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Q&A Assist
    2026-03-30T01:48:00.37+00:00
    AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more
    1. Usage of the Python SDK (openai package)

    The deprecation described applies to the Azure AI Inference beta SDK (azure-ai-inference preview package). The provided context only discusses:

    • Azure OpenAI model lifecycle (deprecations/retirements of models)
    • API version lifecycle at a high level

    There is no statement in the context that the openai Python package (used with Azure OpenAI) is part of the Azure AI Inference beta SDK retirement. Based on the context, only the beta azure-ai-inference SDK is explicitly called out for retirement; no additional impact on the openai package is described.

    1. Usage of API version 2024-10-21 and relation to the SDK retirement date

    From the context:

    • API versions have their own lifecycle, tracked separately in the API version lifecycle article.
    • The guidance is: if using a preview API version to access the latest features, consult the API version lifecycle article to understand how long that version will be supported.

    The context does not state any requirement to change API versions by May 30, 2025 due to SDK retirement. It treats:

    • SDK lifecycle, and
    • API version lifecycle

    as separate concerns. API versions are governed by the API lifecycle documentation, not by the SDK retirement date.

    Therefore, according to the context, API version support is managed independently via the API version lifecycle, and no specific May 30, 2025 deadline for changing api-version is documented.

    1. Impact on Azure OpenAI resources and endpoints (https://[resource].openai.azure.com/)

    From the model retirement and lifecycle documentation in the context:

    • The focus is on model availability, deprecation, and retirement (for example, how long GA models are available, how upgrades occur, and how customers are notified).
    • There is no statement requiring migration from classic Azure OpenAI Service resources to Azure AI Foundry resources.
    • There is no requirement documented to change endpoint formats or recreate resources as part of the model or API lifecycle.

    The documented changes concern:

    • Which models and model versions are available and when they are deprecated/retired.
    • How customers are notified (Azure Resource Health, email) and the minimum availability windows for GA models.

    No requirement is described in the context to:

    • Switch from classic Azure OpenAI resources to Foundry resources,
    • Change endpoint URLs from https://[resource-name].openai.azure.com/, or
    • Recreate existing Azure OpenAI resources solely due to the model/API lifecycle.

    Any impact described in the context is limited to model and API version availability and upgrade behavior, not to the resource type or endpoint scheme.

    References:

    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.