Managed online endpoints SKU list

This table shows the VM SKUs that are supported for Azure Machine Learning managed online endpoints.

• The full SKU names listed in the table can be used for Azure CLI or Azure Resource Manager templates (ARM templates) requests to create and update deployments.

• For more information on configuration details such as CPU and RAM, see Azure Machine Learning Pricing and VM sizes.

Relative Size	General Purpose	Compute Optimized	Memory Optimized	GPU
V.Small	Standard_DS1_v2 Standard_DS2_v2	Standard_F2s_v2	Standard_E2s_v3	Standard_NC4as_T4_v3
Small	Standard_DS3_v2	Standard_F4s_v2	Standard_E4s_v3	Standard_NC6s_v2 Standard_NC6s_v3 Standard_NC8as_T4_v3
Medium	Standard_DS4_v2	Standard_F8s_v2	Standard_E8s_v3	Standard_NC12s_v2 Standard_NC12s_v3 Standard_NC16as_T4_v3
Large	Standard_DS5_v2	Standard_F16s_v2	Standard_E16s_v3	Standard_NC24s_v2 Standard_NC24s_v3 Standard_NC64as_T4_v3
X-Large	-	Standard_F32s_v2 Standard_F48s_v2 Standard_F64s_v2 Standard_F72s_v2 Standard_FX24mds Standard_FX36mds Standard_FX48mds	Standard_E32s_v3 Standard_E48s_v3 Standard_E64s_v3	Standard_ND40rs_v2 Standard_ND96asr_v4 Standard_ND96amsr_A100_v4

Caution

Standard_DS1_v2 and Standard_F2s_v2 may be too small for bigger models and may lead to container termination due to insufficient memory, not enough space on the disk, or probe failure as it takes too long to initiate the container. If you want to reduce the cost of deploying multiple models with managed online endpoint, see the example for multi models. If you face OutOfQuota errors or ReourceNotReady errors, try bigger VM SKUs.