diff --git a/semantic_router/llms/__init__.py b/semantic_router/llms/__init__.py
index 4e2eef16f35d5ef726ed121c6159ab27446ada6a..36f13c8dcb003372cd1bf87112ea5838388ff911 100644
--- a/semantic_router/llms/__init__.py
+++ b/semantic_router/llms/__init__.py
@@ -1,5 +1,6 @@
 from semantic_router.llms.base import BaseLLM
 from semantic_router.llms.cohere import CohereLLM
+from semantic_router.llms.llamacpp import LlamaCppLLM
 from semantic_router.llms.mistral import MistralAILLM
 from semantic_router.llms.openai import OpenAILLM
 from semantic_router.llms.openrouter import OpenRouterLLM
@@ -8,6 +9,7 @@ from semantic_router.llms.zure import AzureOpenAILLM
 __all__ = [
     "BaseLLM",
     "OpenAILLM",
+    "LlamaCppLLM",
     "OpenRouterLLM",
     "CohereLLM",
     "AzureOpenAILLM",
diff --git a/semantic_router/llms/llamacpp.py b/semantic_router/llms/llamacpp.py
index 1126d0558c0bd051f4b39e55b84f3e1c923be086..9a66b732c16b1c0779d60a45c8551f4ebc17e300 100644
--- a/semantic_router/llms/llamacpp.py
+++ b/semantic_router/llms/llamacpp.py
@@ -2,8 +2,6 @@ from contextlib import contextmanager
 from pathlib import Path
 from typing import Any, Optional
 
-# from llama_cpp import Llama, LlamaGrammar
-
 from semantic_router.llms.base import BaseLLM
 from semantic_router.schema import Message
 from semantic_router.utils.logger import logger