diff --git a/llama.cpp b/llama.cpp
index e6cd7e89b..eff3a6bb5 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1077,7 +1077,11 @@ static void llama_model_load_internal(
     #if defined(GGML_USE_CLBLAST)
     if (file_version == LLAMA_FILE_VERSION_GGJT_V3) {
         if (hparams.ftype >= LLAMA_FTYPE_MOSTLY_Q2_K && hparams.ftype <= LLAMA_FTYPE_MOSTLY_Q6_K) {
-            printf("\n===\nK-Quants are currently not supported with CLBlast!!!\nPlease select a q4_0, q4_0, q5_0 or q5_1 format instead!\n=====\n");
+            if(n_gpu_layers>0)
+            {
+                n_gpu_layers = 0;
+                printf("\n===\nCLBlast cannot offload layers for K-Quants!\nPlease select a q4_0, q4_0, q5_0 or q5_1 format instead!\n=====\n");
+            }
         }
     }
     #endif