All Products
Search

This Product

    Platform For AI:Develop and use DeepSeek models

    Document Center

    Platform For AI:Develop and use DeepSeek models

    Last Updated:Jan 21, 2026

    Model Gallery is a model-as-a-service (MaaS) component of PAI that provides state-of-the-art models across large language model (LLM), AI-generated content (AIGC), computer vision (CV), and natural language processing (NLP). This topic describes how to deploy, fine-tune, distill, and build applications with DeepSeek models using Model Gallery.

    Before you start

    (Required) Activate PAI and create a workspace

    PAI workspaces centralize management of computing resources, permissions, and AI assets. Activating PAI creates a default workspace. If none exists, create one manually. OSS is activated by default for storing code, models, and datasets.

    Click to view the procedure

    1. Go to the PAI console to activate PAI. Each region requires separate activation.

      image.png

    2. (Optional) If OSS is not activated, go to the OSS console and create a bucket in the same region as your PAI workspace:

      image.png

    3. If no workspace exists, go to the PAI Workspaces page and click Create Workspace.

      image.png

    4. Go to Model Gallery to find your model.image.png

    (Note) Region and resource specifications

    1. Most Alibaba Cloud services, including PAI workspaces and OSS buckets, are region-specific. Some regions are not interconnected, so select your region carefully.

    2. Computing resource availability varies by region. If resources are unavailable in one region, check other regions.

      image.png

    3. PAI offers pay-as-you-go and subscription billing. Pay-as-you-go resources are shared across users, so stockouts may occur. Check other regions if needed.

    4. PAI offers various resource specifications. Some are restricted to whitelisted users. Consult your sales manager for recommendations.

      PAI also supports Lingjun resources with high-speed networks for distributed training or deployment. These are restricted to whitelisted users. Contact your sales manager if needed.

      image.png

    (Optional) Create a VPC for distributed training or deployment

    A VPC is required for distributed training or deployment. For internet access, configure a public port.

    When creating a VPC, also configure a vSwitch and security group.

    1. Create a VPC and vSwitch.

    2. Create a security group in the same region.

    Specify your VPC when starting training or deployment in Model Gallery. image.png

    Model deployment

    One-click deployment of DeepSeek-V3 and DeepSeek-R1 models.

    For quick practice, try DeepSeek-R1-Distill-Qwen-7B. This smaller distilled model has low resource requirements and can be deployed with free trial resources.

    Fine-tuning and distillation

    Fine-tuning: Train the model with your data to improve accuracy for your use case.

    Distillation: Transfer knowledge from a larger teacher model to a smaller student model, retaining accuracy while reducing compute and storage costs.

    Note

    Fine-tuning success depends on dataset quality, hyperparameters, and experimentation. For many use cases, RAG may be simpler and sufficient.

    One-click fine-tuning of DeepSeek-R1 distill models.

    Build AI application

    Develop applications with LangStudio

    PAI LangStudio simplifies enterprise LLM application development with built-in templates for RAG, web search, and other AI application types.