diff --git a/llama_stack/templates/starter/run.yaml b/llama_stack/templates/starter/run.yaml
index b3dfe32d5..ed1ead03f 100644
--- a/llama_stack/templates/starter/run.yaml
+++ b/llama_stack/templates/starter/run.yaml
@@ -256,6 +256,11 @@ inference_store:
   type: sqlite
   db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/starter}/inference_store.db
 models:
+- metadata:
+    embedding_dimension: 384
+  model_id: all-MiniLM-L6-v2
+  provider_id: ${env.ENABLE_SENTENCE_TRANSFORMERS:=sentence-transformers}
+  model_type: embedding
 - metadata: {}
   model_id: ${env.ENABLE_CEREBRAS:=__disabled__}/llama3.1-8b
   provider_id: ${env.ENABLE_CEREBRAS:=__disabled__}
@@ -1162,11 +1167,6 @@ models:
   provider_id: ${env.ENABLE_SAMBANOVA:=__disabled__}
   provider_model_id: sambanova/Meta-Llama-Guard-3-8B
   model_type: llm
-- metadata:
-    embedding_dimension: 384
-  model_id: all-MiniLM-L6-v2
-  provider_id: ${env.ENABLE_SENTENCE_TRANSFORMERS:=sentence-transformers}
-  model_type: embedding
 shields:
 - shield_id: ${env.ENABLE_OLLAMA:=__disabled__}
   provider_id: llama-guard
diff --git a/llama_stack/templates/starter/starter.py b/llama_stack/templates/starter/starter.py
index 6b8aa8974..b96602abc 100644
--- a/llama_stack/templates/starter/starter.py
+++ b/llama_stack/templates/starter/starter.py
@@ -380,7 +380,7 @@ def get_distribution_template() -> DistributionTemplate:
                     "files": [files_provider],
                     "post_training": [post_training_provider],
                 },
-                default_models=default_models + [embedding_model],
+                default_models=[embedding_model] + default_models,
                 default_tool_groups=default_tool_groups,
                 # TODO: add a way to enable/disable shields on the fly
                 default_shields=shields,