diff --git a/kischdle/llmux/llmux/backends/llamacpp.py b/kischdle/llmux/llmux/backends/llamacpp.py
index f2da464..aae0991 100644
--- a/kischdle/llmux/llmux/backends/llamacpp.py
+++ b/kischdle/llmux/llmux/backends/llamacpp.py
@@ -38,13 +38,10 @@ class LlamaCppBackend(BaseBackend):
         logger.info(f"Loading GGUF model {model_path} with n_gpu_layers={n_gpu_layers}")
 
         def _load():
-            from llama_cpp import GGML_TYPE_Q8_0
             kwargs = {
                 "model_path": model_path,
                 "n_gpu_layers": n_gpu_layers,
                 "n_ctx": 4096,
-                "type_k": GGML_TYPE_Q8_0,
-                "type_v": GGML_TYPE_Q8_0,
                 "verbose": False,
             }
             if physical.mmproj_file: