diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 9ccfc7fc61c47..26ac5e99bfc7a 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -271,19 +271,32 @@ static buft_list_t make_cpu_buft_list(const std::vector<ggml_backend_dev_t> & de
         }
     }
 
-    // add extra buffer types
-    auto * cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
-    auto * cpu_reg = ggml_backend_dev_backend_reg(cpu_dev);
-    auto ggml_backend_dev_get_extra_bufts_fn = (ggml_backend_dev_get_extra_bufts_t)
-        ggml_backend_reg_get_proc_address(cpu_reg, "ggml_backend_dev_get_extra_bufts");
-    if (ggml_backend_dev_get_extra_bufts_fn) {
-        ggml_backend_buffer_type_t * extra_bufts = ggml_backend_dev_get_extra_bufts_fn(cpu_dev);
-        while (extra_bufts && *extra_bufts) {
-            buft_list.emplace_back(cpu_dev, *extra_bufts);
-            ++extra_bufts;
+    bool has_gpu_device = false;
+    for (auto * dev : devices) {
+        if (ggml_backend_dev_type(dev) == GGML_BACKEND_DEVICE_TYPE_GPU) {
+            has_gpu_device = true;
+            break;
         }
     }
 
+    // add extra buffer types, only if no GPU device is present
+    // ref: https://github.com/ggml-org/llama.cpp/issues/12481#issuecomment-2743136094
+    if (!has_gpu_device) {
+        auto * cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
+        auto * cpu_reg = ggml_backend_dev_backend_reg(cpu_dev);
+        auto ggml_backend_dev_get_extra_bufts_fn = (ggml_backend_dev_get_extra_bufts_t)
+            ggml_backend_reg_get_proc_address(cpu_reg, "ggml_backend_dev_get_extra_bufts");
+        if (ggml_backend_dev_get_extra_bufts_fn) {
+            ggml_backend_buffer_type_t * extra_bufts = ggml_backend_dev_get_extra_bufts_fn(cpu_dev);
+            while (extra_bufts && *extra_bufts) {
+                buft_list.emplace_back(cpu_dev, *extra_bufts);
+                ++extra_bufts;
+            }
+        }
+    } else {
+        LLAMA_LOG_WARN("%s: disabling extra buffer types (i.e. repacking) since a GPU device is available\n", __func__);
+    }
+
     // add a host buffer type
     // storing the tensors in a host buffer is useful when the processing of large batches
     // is offloaded to a GPU device, since it reduces the time spent on data transfers