--- a/src/cpp/src/gguf_utils/gguf.hpp
+++ b/src/cpp/src/gguf_utils/gguf.hpp
@@ -28,7 +28,12 @@
                             std::unordered_map<std::string, gguf_tensor_type>>;
 
 template <typename... Args>
-std::string format(std::string fmt, Args... args);
+std::string format(std::string fmt, Args... args) {
+    char buf[1000];
+    int n = snprintf(buf, sizeof(buf), fmt.c_str(), args...);
+    assert(n >= 0 && n < static_cast<int>(sizeof(buf)));
+    return std::string(buf);
+}
 
 ov::Shape get_shape(const gguf_tensor& tensor);
 
--- a/src/cpp/src/gguf_utils/gguf.cpp
+++ b/src/cpp/src/gguf_utils/gguf.cpp
@@ -13,18 +13,6 @@
 // https://github.com/antirez/gguf-tools/blob/af7d88d808a7608a33723fba067036202910acb3/gguflib.h#L102-L108
 constexpr int gguf_array_header_size = 12;
 
-template <typename... Args>
-std::string format(std::string fmt, Args... args) {
-    size_t bufferSize = 1000;
-    char* buffer = new char[bufferSize];
-    int n = sprintf(buffer, fmt.c_str(), args...);
-    assert(n >= 0 && n < (int)bufferSize - 1);
-
-    std::string fmtStr(buffer);
-    delete[] buffer;
-    return fmtStr;
-}
-
 std::optional<uint32_t> dtype_to_gguf_tensor_type(const ov::element::Type& dtype) {
     switch (dtype) {
     case ov::element::f64: