diff --git a/ggml/src/ggml-cpu/ops.cpp b/ggml/src/ggml-cpu/ops.cpp
index 9f1e5f8d64632..e573e78fae122 100644
--- a/ggml/src/ggml-cpu/ops.cpp
+++ b/ggml/src/ggml-cpu/ops.cpp
@@ -9,7 +9,6 @@
 
 #include <cfloat>
 #include <algorithm>
-#include <functional>
 
 // ggml_compute_forward_dup
 
@@ -7665,6 +7664,18 @@ void ggml_compute_forward_timestep_embedding(
 
 // ggml_compute_forward_argsort
 
+template<ggml_sort_order order>
+struct cmp_argsort {
+    const float * data;
+    bool operator()(int32_t a, int32_t b) const {
+        if constexpr (order == GGML_SORT_ORDER_ASC) {
+            return data[a] < data[b];
+        } else {
+            return data[a] > data[b];
+        }
+    }
+};
+
 static void ggml_compute_forward_argsort_f32(
     const ggml_compute_params * params,
     ggml_tensor * dst) {
@@ -7691,16 +7702,17 @@ static void ggml_compute_forward_argsort_f32(
             dst_data[j] = j;
         }
 
-        std::function<bool(int32_t, int32_t)> cmp;
-
-        // note: this might be causing memory allocations? ideally should be avoided if it's the case
         switch (order) {
-            case GGML_SORT_ORDER_ASC:  cmp = [src_data](int32_t a, int32_t b) { return src_data[a] < src_data[b]; }; break;
-            case GGML_SORT_ORDER_DESC: cmp = [src_data](int32_t a, int32_t b) { return src_data[a] > src_data[b]; }; break;
+            case GGML_SORT_ORDER_ASC:
+                std::sort(dst_data, dst_data + ne0, cmp_argsort<GGML_SORT_ORDER_ASC>{src_data});
+                break;
+
+            case GGML_SORT_ORDER_DESC:
+                std::sort(dst_data, dst_data + ne0, cmp_argsort<GGML_SORT_ORDER_DESC>{src_data});
+                break;
+
             default: GGML_ABORT("invalid sort order");
         }
-
-        std::sort(dst_data, dst_data + ne0, cmp);
     }
 }