diff --git a/docs/ramalama-serve.1.md b/docs/ramalama-serve.1.md
index 4181e5be..12dabbac 100644
--- a/docs/ramalama-serve.1.md
+++ b/docs/ramalama-serve.1.md
@@ -79,7 +79,7 @@ After=local-fs.target
 [Container]
 AddDevice=-/dev/dri
 AddDevice=-/dev/kfd
-Exec=llama-server --port 1234 -m $HOME/.local/share/ramalama/models/huggingface/instructlab/granite-7b-lab-GGUF/granite-7b-lab-Q4_K_M.gguf
+Exec=python3 -m llama_cpp.server --port 1234 --model $HOME/.local/share/ramalama/models/huggingface/instructlab/granite-7b-lab-GGUF/granite-7b-lab-Q4_K_M.gguf
 Image=quay.io/ramalama/ramalama:latest
 Mount=type=bind,src=/home/dwalsh/.local/share/ramalama/models/huggingface/instructlab/granite-7b-lab-GGUF/granite-7b-lab-Q4_K_M.gguf,target=/mnt/models/model.file,ro,Z
 ContainerName=MyGraniteServer
diff --git a/ramalama/cli.py b/ramalama/cli.py
index 2d13ec58..77b0fc9a 100644
--- a/ramalama/cli.py
+++ b/ramalama/cli.py
@@ -105,7 +105,7 @@ def load_and_merge_config():
         )
 
     ramalama_config['carimage'] = ramalama_config.get('carimage', "registry.access.redhat.com/ubi9-micro:latest")
-    ramalama_config['runtime'] = ramalama_config.get('runtime', 'llama.cpp')
+    ramalama_config['runtime'] = ramalama_config.get('runtime', 'llama-cpp-python')
     ramalama_config['store'] = os.getenv('RAMALAMA_STORE', ramalama_config.get('store', get_store()))
     ramalama_config['transport'] = os.getenv('RAMALAMA_TRANSPORT', ramalama_config.get('transport', "ollama"))
 
@@ -207,8 +207,8 @@ def configure_arguments(parser):
     parser.add_argument(
         "--runtime",
         default=config.get("runtime"),
-        choices=["llama.cpp", "vllm"],
-        help="specify the runtime to use; valid options are 'llama.cpp' and 'vllm'",
+        choices=["llama-cpp-python", "llama.cpp", "vllm"],
+        help="specify the runtime to use; valid options are 'llama-cpp-python', 'llama.cpp' and 'vllm'",
     )
     parser.add_argument(
         "--store",
diff --git a/ramalama/model.py b/ramalama/model.py
index c8d6dcae..c21ef404 100644
--- a/ramalama/model.py
+++ b/ramalama/model.py
@@ -284,6 +284,19 @@ def run(self, args):
                 raise NotImplementedError(file_not_found_in_container % (exec_args[0], str(e).strip("'")))
             raise NotImplementedError(file_not_found % (exec_args[0], exec_args[0], exec_args[0], str(e).strip("'")))
 
+    def execute_model(self, model_path, exec_args, args):
+        try:
+            if self.exec_model_in_container(model_path, exec_args, args):
+                return
+            if args.dryrun:
+                dry_run(exec_args)
+                return
+            exec_cmd(exec_args, debug=args.debug)
+        except FileNotFoundError as e:
+            if in_container():
+                raise NotImplementedError(file_not_found_in_container % (exec_args[0], str(e).strip("'")))
+            raise NotImplementedError(file_not_found % (exec_args[0], exec_args[0], exec_args[0], str(e).strip("'")))
+
     def serve(self, args):
         if hasattr(args, "name") and args.name:
             if not args.container and not args.generate:
@@ -300,13 +313,25 @@ def serve(self, args):
         if not args.container and not args.generate:
             exec_model_path = model_path
 
-        exec_args = ["llama-server", "--port", args.port, "-m", exec_model_path]
         if args.runtime == "vllm":
             exec_args = ["vllm", "serve", "--port", args.port, exec_model_path]
-        else:
+        elif args.runtime == "llama.cpp":
+            exec_args = ["llama-server", "--port", args.port, "-m", exec_model_path, "--host", args.host]
             if args.gpu:
                 exec_args.extend(self.gpu_args())
-            exec_args.extend(["--host", args.host])
+
+        else:
+            exec_args = [
+                "python3",
+                "-m",
+                "llama_cpp.server",
+                "--port",
+                args.port,
+                "--model",
+                exec_model_path,
+                "--host",
+                args.host,
+            ]
 
         if args.generate == "quadlet":
             return self.quadlet(model_path, args, exec_args)
@@ -317,17 +342,7 @@ def serve(self, args):
         if args.generate == "quadlet/kube":
             return self.quadlet_kube(model_path, args, exec_args)
 
-        try:
-            if self.exec_model_in_container(model_path, exec_args, args):
-                return
-            if args.dryrun:
-                dry_run(exec_args)
-                return
-            exec_cmd(exec_args, debug=args.debug)
-        except FileNotFoundError as e:
-            if in_container():
-                raise NotImplementedError(file_not_found_in_container % (exec_args[0], str(e).strip("'")))
-            raise NotImplementedError(file_not_found % (exec_args[0], exec_args[0], exec_args[0], str(e).strip("'")))
+        self.execute_model(model_path, exec_args, args)
 
     def quadlet(self, model, args, exec_args):
         quadlet = Quadlet(model, args, exec_args)
diff --git a/test/system/040-serve.bats b/test/system/040-serve.bats
index 5c2b4f3d..c9c77eaa 100644
--- a/test/system/040-serve.bats
+++ b/test/system/040-serve.bats
@@ -136,7 +136,7 @@ verify_begin=".*run --rm -i --label RAMALAMA --security-opt=label=disable --name
 
     run cat tinyllama.container
     is "$output" ".*PublishPort=1234" "PublishPort should match"
-    is "$output" ".*Exec=llama-server --port 1234 -m .*" "Exec line should be correct"
+    is "$output" ".*Exec=python3 -m llama_cpp.server --port 1234 --model .*" "Exec line should be correct"
     is "$output" ".*Mount=type=bind,.*tinyllama" "Mount line should be correct"
 
     rm tinyllama.container
@@ -174,7 +174,7 @@ verify_begin=".*run --rm -i --label RAMALAMA --security-opt=label=disable --name
 	run cat $name.container
 	is "$output" ".*PublishPort=1234" "PublishPort should match"
 	is "$output" ".*ContainerName=${name}" "Quadlet should have ContainerName field"
-	is "$output" ".*Exec=llama-server --port 1234 -m .*" "Exec line should be correct"
+	is "$output" ".*Exec=python3 -m llama_cpp.server --port 1234 --model .*" "Exec line should be correct"
 	is "$output" ".*Mount=type=image,source=${ociimage},destination=/mnt/models,subpath=/models,readwrite=false" "Volume line should be correct"
 
 	if is_container; then
@@ -226,7 +226,7 @@ verify_begin=".*run --rm -i --label RAMALAMA --security-opt=label=disable --name
 
     run cat $name.yaml
     is "$output" ".*image: quay.io/ramalama/ramalama:latest" "Should container image"
-    is "$output" ".*command: \[\"llama-server\"\]" "Should command"
+    is "$output" ".*command: \[\"python3\"\]" "Should command"
     is "$output" ".*containerPort: 1234" "Should container container port"
 
     run_ramalama serve --name=${name} --port 1234 --generate=quadlet/kube ${model}
@@ -235,7 +235,7 @@ verify_begin=".*run --rm -i --label RAMALAMA --security-opt=label=disable --name
 
     run cat $name.yaml
     is "$output" ".*image: quay.io/ramalama/ramalama:latest" "Should container image"
-    is "$output" ".*command: \[\"llama-server\"\]" "Should command"
+    is "$output" ".*command: \[\"llama_cpp.server\"\]" "Should command"
     is "$output" ".*containerPort: 1234" "Should container container port"
 
     run cat $name.kube
diff --git a/test/system/060-info.bats b/test/system/060-info.bats
index 44bf3f91..de4d29cc 100644
--- a/test/system/060-info.bats
+++ b/test/system/060-info.bats
@@ -17,7 +17,7 @@ load helpers
     # FIXME Engine  (podman|docker|'')
     tests="
 Image   | "quay.io/ramalama/ramalama:latest"
-Runtime | "llama.cpp"
+Runtime | "llama-cpp-python"
 Version | "${version}"
 Store   | \\\("${HOME}/.local/share/ramalama"\\\|"/var/lib/ramalama"\\\)
 "