Choose Your Platform

Smart Studio offers two deployment paths. Choose the one that fits your use case:

• BYO-GPU (Model Serving Platform) — Deploy open-source models on your own Compute Cards for internal use or integration into an existing platform.
• BYO-Key (API Router Platform) — Resell Model APIs through a white-labelled User Console and Admin Console.

Both paths share the same activation workflow: choose a billing model, install the agent, verify your machines, and deploy the platform.

SMART STUDIO DOCS 2.0

Start with the outcome you need

Choose one maintained path. Each path connects the prerequisite, Console action, successful result, recovery guidance, and next step.

New in 2.0Choose your platform: BYO-GPU to deploy models on your own hardware, or BYO-Key to resell Model APIs.

MODEL SERVICE PLATFORM

Activate a GPU cluster

Result: Your cluster is Activated, its expected machines are visible, and its GPU capacity is ready for model deployment.

Start with requirements→

1. 1Check requirementsConfirm supported infrastructure, network access, and operator permissions.Prerequisite
2. 2Choose the billing modelSelect Model Service Platform and confirm the billing model (Token Sharing or Fixed Amount Fee).Decision
3. 3Activate the clusterInstall the platform agent, submit machine configuration, and complete the guided checks.Core action
4. 4Verify machines and capacityConfirm the activation state, expected nodes, available Compute Cards, and resource health.Verify
5. 5Deploy a modelChoose a platform model in Model Gallery or upload your own, then create a deployment.Next step

WHAT SUCCESS LOOKS LIKE

Result checkVerified

• ✓Activation shows Activated
• ✓Expected nodes are visible
• ✓GPU capacity is healthy

Do not continue until the expected result is visible in the Console.

Next recommended pageGPU ObservabilityVerify your capacity →

Platform Comparison

The following table compares the two platforms:

Comparison	API Router Platform	Model Serving Platform
Positioning	Model API resale	Model serving and management
Suitable scenarios	Resell model APIs through branded customer-facing consoles.	Deploy and serve proprietary models for internal use or integrate model services into existing platforms for monetization.
Console / usage	Admin Console for platform operators: model listing, pricing, branding, customer management, bring your own key (BYOK), analytics, and gateway monitoring. User Console for end users: model playground, API invocation, API key management, and account top-up.	A single Model Serving Platform for managing proprietary models, model deployments, training and evaluation, cluster resources, and model service health.
Key capabilities	- Model listing - Pricing configuration - White-labeling - Admin user management - Customer management - Bring your own key (BYOK) - Usage analytics - Gateway monitoring - More capabilities	- Proprietary model management - Inference-optimized model hosting - Model deployment with version management - Dataset management - Model training and evaluation - Cluster monitoring - Model service monitoring - More capabilities
Learn more	View API Router Platform overview	View Model Serving Platform overview

Before You Start

You need:

• An Alibaba Cloud account.
• A CPU server that can reach the internet during activation.
• One or more supported GPU servers that meet the BYO-GPU requirements.
• SSH access and the network details required by the activation wizard.

The Console is the product map

Docs 2.0 follows the current public Console menu. If a capability is not available in the Console, it is not presented as a supported 2.0 workflow.

Choose Your Next Step

New to Smart Studio

Start with Activate Cluster. It covers billing model selection, agent installation, machine verification, platform deployment, and recovery from common failures.

Your cluster is already activated

Open GPU Observability to confirm resource health, then go to Model Gallery or My Models.

You already have a model

Follow Create Deployment, then use Manage Deployments to verify its state and manage its lifecycle.

Where Each Console Menu Leads

Console menu	Use it for
Activate Cluster	Choose a billing model, connect GPU infrastructure, and deploy the platform.
Model Gallery	Discover platform models and continue to deployment.
My Models	Upload and manage custom or fine-tuned model assets.
Deployments	Create, monitor, stop, restart, edit, or delete deployments.
Datasets / Fine-Tuning / Evaluations	Prepare data, improve models, and measure quality.
Usage	Review consumption by self-deployed workloads.
Billing	Review charges and billing details.