Add a model to a provider device

Choose an eligible model#

Open the model catalog from your Provider workflow or application. Search for the intended model and read the exact identifier, hardware requirements, serving permissions and available commercial terms. A recognizable model name is not enough to identify its exact tag or variant.

Check device capacity#

GPU memory needs depend on weights, precision or quantization, context length, runtime overhead and concurrency. Use the current model requirements and observed device resources. Do not infer guaranteed compatibility from parameter count alone.

Install and enable#

1. Select the exact model and review any access or pricing terms.
2. Start the supported model download or connection flow.
3. Wait for installation or synchronization to finish and resolve errors before enabling sharing.
4. Enable the model for sharing on the intended device.
5. Confirm the model and device state in Console, then monitor real activity.

When a model does not appear#

Check the selected workspace, provider approval, catalog permissions, model availability and whether the installation finished. For automated catalog checks, use Provider model endpoints.

When a model cannot load#

Read the runtime error, compare required and available resources and check whether other workloads consume the GPU. Resolve loading problems before advertising capacity. Changing catalog text does not fix a device runtime failure.