diff --git a/backend/Dockerfile.model_server b/backend/Dockerfile.model_server
index f2345be9ef..1c959e8ed6 100644
--- a/backend/Dockerfile.model_server
+++ b/backend/Dockerfile.model_server
@@ -82,4 +82,4 @@ ENV PYTHONPATH=/app
 ARG ONYX_VERSION=0.0.0-dev
 ENV ONYX_VERSION=${ONYX_VERSION}
 
-CMD ["uvicorn", "model_server.main:app", "--host", "0.0.0.0", "--port", "9000"]
+CMD ["python", "-m", "model_server"]
diff --git a/backend/model_server/__main__.py b/backend/model_server/__main__.py
new file mode 100644
index 0000000000..52780a487c
--- /dev/null
+++ b/backend/model_server/__main__.py
@@ -0,0 +1,29 @@
+"""Process entry point for the model server (`python -m model_server`).
+
+The `DISABLE_MODEL_SERVER` gate runs here, ahead of `model_server.main`'s heavy ML
+imports, so a disabled container exits without loading torch / the model stack.
+"""
+
+import sys
+
+from onyx.utils.logger import setup_logger
+from shared_configs.configs import DISABLE_MODEL_SERVER
+
+logger = setup_logger()
+
+
+def main() -> None:
+    if DISABLE_MODEL_SERVER:
+        # The deployment points inference/indexing at an external model server, so this
+        # container has nothing to run. Exit cleanly instead of starting uvicorn.
+        logger.notice("DISABLE_MODEL_SERVER is set; skipping model server startup.")
+        sys.exit(0)
+
+    # Imported lazily so the disabled path above stays free of the heavy ML imports.
+    from model_server.main import run_server
+
+    run_server()
+
+
+if __name__ == "__main__":
+    main()
diff --git a/backend/model_server/main.py b/backend/model_server/main.py
index b8c80a38f2..cfcd7df11b 100644
--- a/backend/model_server/main.py
+++ b/backend/model_server/main.py
@@ -24,7 +24,6 @@ from onyx.utils.middleware import add_onyx_request_id_middleware
 from onyx.utils.middleware import add_onyx_tenant_id_middleware
 from shared_configs.configs import INDEXING_ONLY
 from shared_configs.configs import MIN_THREADS_ML_MODELS
-from shared_configs.configs import MODEL_SERVER_ALLOWED_HOST
 from shared_configs.configs import MODEL_SERVER_PORT
 from shared_configs.configs import SENTRY_DSN
 from shared_configs.configs import SENTRY_TRACES_SAMPLE_RATE
@@ -141,11 +140,20 @@ def get_model_app() -> FastAPI:
 app = get_model_app()
 
 
-if __name__ == "__main__":
+def run_server() -> None:
+    # Bind all interfaces (loopback included) so the container healthcheck's
+    # localhost probe reaches the server. Mirrors the old shell command's
+    # `--host 0.0.0.0`; MODEL_SERVER_HOST is a client-side address and must not
+    # drive the bind host.
+    host = "0.0.0.0"  # noqa: S104
     logger.notice(
         "Starting Onyx Model Server on http://%s:%s/",
-        MODEL_SERVER_ALLOWED_HOST,
+        host,
         str(MODEL_SERVER_PORT),
     )
     logger.notice("Model Server Version: %s", __version__)
-    uvicorn.run(app, host=MODEL_SERVER_ALLOWED_HOST, port=MODEL_SERVER_PORT)
+    uvicorn.run(app, host=host, port=MODEL_SERVER_PORT)
+
+
+if __name__ == "__main__":
+    run_server()
diff --git a/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_indexing_service_template.yaml b/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_indexing_service_template.yaml
index b1975fac9a..580c17bb14 100644
--- a/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_indexing_service_template.yaml
+++ b/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_indexing_service_template.yaml
@@ -120,12 +120,6 @@ Resources:
           Image: onyxdotapp/onyx-model-server:latest
           Cpu: 0
           Essential: true
-          Command:
-            - "/bin/sh"
-            - "-c"
-            - >
-              if [ "${DISABLE_MODEL_SERVER}" = "True" ] || [ "${DISABLE_MODEL_SERVER}" = "true" ]; then echo 'Skipping service...';
-              exit 0; else exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000; fi
           PortMappings:
             - Name: model-server
               ContainerPort: 9000
diff --git a/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_inference_service_template.yaml b/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_inference_service_template.yaml
index 98658aa4c5..5256a60a65 100644
--- a/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_inference_service_template.yaml
+++ b/deployment/aws_ecs_fargate/cloudformation/services/onyx_model_server_inference_service_template.yaml
@@ -120,12 +120,6 @@ Resources:
           Image: onyxdotapp/onyx-model-server:latest
           Cpu: 0
           Essential: true
-          Command:
-            - "/bin/sh"
-            - "-c"
-            - >
-              if [ "${DISABLE_MODEL_SERVER}" = "True" ] || [ "${DISABLE_MODEL_SERVER}" = "true" ]; then echo 'Skipping service...';
-              exit 0; else exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000; fi
           PortMappings:
             - Name: model-server
               ContainerPort: 9000
diff --git a/deployment/docker_compose/docker-compose.multitenant-dev.yml b/deployment/docker_compose/docker-compose.multitenant-dev.yml
index d8d6a4e258..60404bb434 100644
--- a/deployment/docker_compose/docker-compose.multitenant-dev.yml
+++ b/deployment/docker_compose/docker-compose.multitenant-dev.yml
@@ -350,13 +350,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
@@ -379,13 +372,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - INDEX_BATCH_SIZE=${INDEX_BATCH_SIZE:-}
diff --git a/deployment/docker_compose/docker-compose.prod-cloud.yml b/deployment/docker_compose/docker-compose.prod-cloud.yml
index 6938b18c74..ce52a306f2 100644
--- a/deployment/docker_compose/docker-compose.prod-cloud.yml
+++ b/deployment/docker_compose/docker-compose.prod-cloud.yml
@@ -164,13 +164,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
@@ -190,13 +183,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
diff --git a/deployment/docker_compose/docker-compose.prod-no-letsencrypt.yml b/deployment/docker_compose/docker-compose.prod-no-letsencrypt.yml
index 7aeec49176..6c8d17efda 100644
--- a/deployment/docker_compose/docker-compose.prod-no-letsencrypt.yml
+++ b/deployment/docker_compose/docker-compose.prod-no-letsencrypt.yml
@@ -165,13 +165,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
@@ -193,13 +186,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
diff --git a/deployment/docker_compose/docker-compose.prod.yml b/deployment/docker_compose/docker-compose.prod.yml
index e99706953a..df41829c5c 100644
--- a/deployment/docker_compose/docker-compose.prod.yml
+++ b/deployment/docker_compose/docker-compose.prod.yml
@@ -201,13 +201,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: unless-stopped
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
@@ -229,13 +222,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: unless-stopped
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
diff --git a/deployment/docker_compose/docker-compose.search-testing.yml b/deployment/docker_compose/docker-compose.search-testing.yml
index 17430ce065..227d55b198 100644
--- a/deployment/docker_compose/docker-compose.search-testing.yml
+++ b/deployment/docker_compose/docker-compose.search-testing.yml
@@ -113,13 +113,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
@@ -137,13 +130,6 @@ services:
     build:
       context: ../../backend
       dockerfile: Dockerfile.model_server
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     restart: on-failure
     environment:
       - MIN_THREADS_ML_MODELS=${MIN_THREADS_ML_MODELS:-}
diff --git a/deployment/docker_compose/docker-compose.yml b/deployment/docker_compose/docker-compose.yml
index f318bf2e97..6b279ac4ad 100644
--- a/deployment/docker_compose/docker-compose.yml
+++ b/deployment/docker_compose/docker-compose.yml
@@ -274,13 +274,6 @@ services:
     #         - driver: nvidia
     #           count: all
     #           capabilities: [gpu]
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     env_file:
       - path: .env
         required: false
@@ -324,13 +317,6 @@ services:
     #         - driver: nvidia
     #           count: all
     #           capabilities: [gpu]
-    command: >
-      /bin/sh -c "if [ \"${DISABLE_MODEL_SERVER:-}\" = \"True\" ] || [ \"${DISABLE_MODEL_SERVER:-}\" = \"true\" ]; then
-        echo 'Skipping service...';
-        exit 0;
-      else
-        exec uvicorn model_server.main:app --host 0.0.0.0 --port 9000;
-      fi"
     env_file:
       - path: .env
         required: false