diff --git a/llama.cpp b/llama.cpp
index 3d13f89c4dc1a..3701ca16dd1f0 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -9,6 +9,7 @@
 #include <queue>
 #include <regex>
 #include <cassert>
+#include <cstring>
 
 // determine number of model parts based on the dimension
 static const std::unordered_map<int, int> LLAMA_N_PARTS = {