diff --git a/llama.cpp b/llama.cpp
index 6e7a53407..e79251194 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -6498,8 +6498,7 @@ struct llama_context * llama_new_context_with_model(
         && (model->ftype == LLAMA_FTYPE_ALL_F32
             || model->ftype == LLAMA_FTYPE_MOSTLY_F16
             || model->ftype == LLAMA_FTYPE_MOSTLY_Q4_0
-            || model->ftype == LLAMA_FTYPE_MOSTLY_Q4_1
-            || model->ftype == LLAMA_FTYPE_MOSTLY_Q6_K)) {
+            || model->ftype == LLAMA_FTYPE_MOSTLY_Q4_1)) {
         // this allocates all Vulkan resources and memory buffers
         ctx->ctx_kompute = ggml_vk_init();