diff --git a/src/plugins/intel_cpu/src/nodes/kernels/scaled_attn/mha_single_token.cpp b/src/plugins/intel_cpu/src/nodes/kernels/scaled_attn/mha_single_token.cpp
index 5018c9314b42c3..82018ed43c25a3 100644
--- a/src/plugins/intel_cpu/src/nodes/kernels/scaled_attn/mha_single_token.cpp
+++ b/src/plugins/intel_cpu/src/nodes/kernels/scaled_attn/mha_single_token.cpp
@@ -77,6 +77,12 @@ void cvt_copy(TA* dst, TB* src, size_t n) {
         i += inc;
     }
 #else
+    if (std::is_same<TA, float>::value && std::is_same<TB, float>::value) {
+        for (; i + vec_len_f32_neon <= n; i += vec_len_f32_neon) {
+            float32x4_t vb1 = __vld1q_f32(src + i);
+            __vst1q_f32(dst + i, vb1);
+        }
+    }
 #if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC)
     if (std::is_same<TA, ov::float16>::value && std::is_same<TB, ov::float16>::value) {
         for (; i + vec_len_f16_neon <= n; i += vec_len_f16_neon) {
@@ -84,13 +90,6 @@ void cvt_copy(TA* dst, TB* src, size_t n) {
             vst1q_f16(reinterpret_cast<float16_t*>(dst + i), vb1);
         }
     }
-#else
-    if (std::is_same<TA, float>::value && std::is_same<TB, float>::value) {
-        for (; i + vec_len_f32_neon <= n; i += vec_len_f32_neon) {
-            float32x4_t vb1 = __vld1q_f32(src + i);
-            __vst1q_f32(dst + i, vb1);
-        }
-    }
 #endif
 #endif
 #endif