Understand Model Lifecycle, Deprecation & Retirement - Model Studio - Alibaba Cloud_Alibaba Cloud Model Studio(Model Studio) - Alibaba Cloud Model Studio

To optimize resources and provide users with the latest, most advanced models, Alibaba Cloud Model Studio periodically retires legacy models. This topic describes the model retirement process.

Model types

Model Studio offers two types of models: stable and preview. We recommend using stable models for production workloads.

Stable models: Production-ready, commercially supported releases. A model series is available in three forms: the series name (for example, qwen-max, qwen-plus, qwen-turbo), a dated version name (for example, qwen-max-2025-01-25), and a -latest alias (for example, qwen-max-latest). All three forms are stable models.
Preview models: Public preview releases, identified by preview, beta, or experimental in the model ID, or explicitly labeled "Preview" in the documentation or console. Preview models are intended for validating new capabilities and collecting feedback. They are updated more frequently and may be retired later.

Retirement checkpoints

Model retirement means that the model service is discontinued and you must switch to another model to continue using the service. The retirement process includes the following checkpoints:

Checkpoint	Description	Timing
Start of retirement	A retirement announcement is published to inform you of the plan. From this date, new users cannot access the model, and the model's default rate limits (QPM and TPM) may be gradually reduced	Same for stable and preview models
End of Marketing (EoM)	All new purchases of the model stop. For pay-as-you-go models, new customers can no longer subscribe to the model's pay-as-you-go billing service	Stable models: 7 days after the start of retirement; preview models: not applicable
End of Service (EoS)	All model services stop. Existing inference services can no longer be called, console features and documentation are retired, and related data is cleaned up and access permissions are revoked	Stable models: 3 months after the start of retirement; preview models: 14 days after the start of retirement

Notification process

Notification schedule

Stable models: A retirement notice is published no later than 3 months before End of Service (EoS).
Preview models: A retirement notice is published no later than 14 days before End of Service (EoS).

Notification channels

We send notifications via official website announcements, email, and internal messages.

If you have called a model scheduled for retirement in the last 3 months, we will also notify you by email or internal message.

Retirement impact

Starting from the start of retirement: New users cannot access the model, and the QPM (queries per minute) and TPM (tokens per minute) for the retiring model will gradually decrease. For models that received a quota increase, their limits will revert to the default rate limit before this reduction begins. Throughout this period, the model API and related console features will remain fully functional.
Starting from EoM (stable models only): New purchases of the model stop. Under pay-as-you-go billing, new customers cannot subscribe to the model's billing service.
Starting from EoS:
- Model inference: Model inference will be discontinued. API calls to the retired model will fail.
- Console features and official documentation: Associated console features (such as Model Gallery and Experience Center) and the official documentation will also be retired. Related data is cleaned up and access permissions are revoked.

Changes to retirement plans

If risks arise, a retirement may be paused, postponed, or canceled. If a retirement is canceled, the model returns to normal sale and may be included in a later retirement plan.
Published announcements may be updated, for example, to change replacement models or to cancel the retirement of some models. Any update includes a "Change notes" section that describes what changed, why, and how it compares with the previous version.

Actions

Go to the Singapore region's Model Monitoring page to check if your account is using any models scheduled for retirement.

To use models in the China (Beijing) region, go to the China (Beijing) region's Model Monitoring page.
If you use an affected model, test the replacement model's business performance before switching to it.

Deprecated models

Scheduled for deprecation on January 19, 2027

For details, see https://www.alibabacloud.com/zh/notice/detail?_p_lc=1&id=2217.

Deprecated on October 10, 2026

For details, see the following announcements:

Retired on May 31, 2026

For details, see https://www.alibabacloud.com/notice/detail?_p_lc=1&id=1873.

Retired on May 13, 2026

For details, see https://www.alibabacloud.com/notice/detail?_p_lc=1&id=1840.

Deprecated on January 30, 2026

No dedicated announcement was published for this batch.

Deprecated on August 20, 2025

No dedicated announcement was published for this batch.