diff --git a/docs/tutorials/foundation-model-tabular.md b/docs/tutorials/foundation-model-tabular.md
index b514224..302d554 100644
--- a/docs/tutorials/foundation-model-tabular.md
+++ b/docs/tutorials/foundation-model-tabular.md
@@ -135,7 +135,7 @@ predictions = future.result()  # blocks until the job finishes, returns a Series
 
 ## Real-time inference
 
-Deploy the model to a SageMaker endpoint with {py:meth}`~autogluon.cloud.TabularFoundationModel.deploy`, then send requests through the returned {py:class}`~autogluon.cloud.TabularEndpoint`. Pick an `instance_type` based on cost and latency requirements (defaults to `ml.g5.xlarge`):
+Deploy the model to a SageMaker endpoint with {py:meth}`~autogluon.cloud.TabularFoundationModel.deploy`, then send requests through the returned {py:class}`~autogluon.cloud.TabularEndpoint`. Pick an `instance_type` based on cost and latency requirements (defaults to `ml.m5.4xlarge` for Mitra v1 and `ml.g5.xlarge` for the other models):
 
 ```python
 endpoint = model.deploy(instance_type="ml.m5.4xlarge")  # takes a few minutes
diff --git a/src/autogluon/cloud/model/registry.py b/src/autogluon/cloud/model/registry.py
index 041951f..9862435 100644
--- a/src/autogluon/cloud/model/registry.py
+++ b/src/autogluon/cloud/model/registry.py
@@ -102,6 +102,7 @@ class FoundationModelConfig:
         model_source_hyperparameter="hf_cls_model",
         inference_hyperparameters={"fine_tune": False},
         predict_instance_type="ml.m5.4xlarge",
+        deploy_instance_type="ml.m5.4xlarge",
     ),
     "mitra-regressor": FoundationModelConfig(
         problem_type="regression",
@@ -110,6 +111,7 @@ class FoundationModelConfig:
         model_source_hyperparameter="hf_reg_model",
         inference_hyperparameters={"fine_tune": False},
         predict_instance_type="ml.m5.4xlarge",
+        deploy_instance_type="ml.m5.4xlarge",
     ),
     "mitra-v2-classifier": FoundationModelConfig(
         problem_type="multiclass",