How Do I Redeploy My Models After Compute Resource-based Model Services Are Discontinued?
Background
Due to business strategy adjustments, the platform is shifting its focus to enhancing serverless model services. Consequently, compute resource-based model services will be officially discontinued at 24:00 (GMT+08:00) on August 18, 2026. Notifications regarding this discontinuation will be sent via SMS, internal messages, and email.
Affected Region
CN-Hong Kong
Affected Functions
New compute resource-based model services can no longer be deployed. API services for existing model services will become unavailable (for example, model calls will fail, or active service processes will be interrupted). Capacity expansion and maintenance will no longer be supported for existing model services.
Model Redeployment
Once the compute resource-based model services are discontinued, you can seamlessly redeploy models on ModelArts. If you encounter any issues, submit a service ticket on the console.
- Models Supported by ModelArts
Text generation: Qwen3-235B, Qwen3-32B-64k, and GPT-OSS 120B
For details about the models, go to Asset Management > Models.
- Region Supported Model Deployment
CN-Hong Kong
- Model Deployment on ModelArts
Choose Model Deployment > Real-Time Services on ModelArts to deploy a built-in or custom model. For details, see Deploying and Using Models for Real-Time Inference.
FAQs
Q1: Can deployed models be smoothly migrated?
No. If the model is available for one-click deployment on ModelArts, you can directly deploy it on ModelArts. If the model is not available for one-click deployment on ModelArts, you are advised to deploy it as a custom model.
Q2: Do I need to purchase resources again for ModelArts?
You do not need to purchase new resources. You can directly deploy models using existing resources. The procedure is as follows:
- Go to the ModelArts Resource Pools, click
on the right of a resource pool, and select Set Job Type. 
- On the Set Job Type page, set Job Type to X-Infer.

Q3: Will I be additionally charged for using models on ModelArts?
- If you use one-click deployment on ModelArts, you will not be additionally charged.
- If you use the custom deployment on ModelArts, you will be additionally charged for using OBS or SFS.
- If you use LTS, you will be charged for using it.
- If you build additional inference links (to lift the shared link limit for public network access), you will be charged for using them.
For details about the billing, see Billing Items for Inference Deployment (New Edition).
What is your overall rating for this page?
Thank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot