diff --git a/ggml-sycl.cpp b/ggml-sycl.cpp
index db3c24f60..9c9d6cee2 100644
--- a/ggml-sycl.cpp
+++ b/ggml-sycl.cpp
@@ -16022,6 +16022,17 @@ catch (sycl::exception const &exc) {
   std::exit(1);
 }
 
+GGML_CALL void ggml_backend_sycl_get_free_device_memory(int device, size_t *free) try {
+    GGML_SYCL_DEBUG("[SYCL] call ggml_backend_sycl_get_free_device_memory\n");
+    size_t total;
+    ggml_backend_sycl_get_device_memory(device, free, &total);
+}
+catch (sycl::exception const &exc) {
+  std::cerr << exc.what() << "Exception caught at file:" << __FILE__
+            << ", line:" << __LINE__ << std::endl;
+  std::exit(1);
+}
+
 ////////////////////////////////////////////////////////////////////////////////
 
 // backend interface
diff --git a/ggml-sycl.h b/ggml-sycl.h
index a9f776fc1..2bb76f94d 100644
--- a/ggml-sycl.h
+++ b/ggml-sycl.h
@@ -33,6 +33,7 @@ GGML_API GGML_CALL void   ggml_sycl_get_gpu_list(int *id_list, int max_len);
 GGML_API GGML_CALL void   ggml_sycl_get_device_description(int device, char *description, size_t description_size);
 GGML_API GGML_CALL int   ggml_backend_sycl_get_device_count();
 GGML_API GGML_CALL void ggml_backend_sycl_get_device_memory(int device, size_t *free, size_t *total);
+GGML_API GGML_CALL void ggml_backend_sycl_get_free_device_memory(int device, size_t *free);
 GGML_API GGML_CALL int ggml_backend_sycl_get_device_index(int device_id);
 
 // TODO: these are temporary
diff --git a/ggml-vulkan.cpp b/ggml-vulkan.cpp
index 1736ab736..1343260be 100644
--- a/ggml-vulkan.cpp
+++ b/ggml-vulkan.cpp
@@ -5781,6 +5781,21 @@ GGML_CALL void ggml_backend_vk_get_device_memory(int device, size_t * free, size
     }
 }
 
+GGML_CALL void ggml_backend_vk_get_free_device_memory(int device, size_t * free) {
+    GGML_ASSERT(device < (int) vk_instance.device_indices.size());
+
+    vk::PhysicalDevice vkdev = vk_instance.instance.enumeratePhysicalDevices()[vk_instance.device_indices[device]];
+
+    vk::PhysicalDeviceMemoryProperties memprops = vkdev.getMemoryProperties();
+
+    for (const vk::MemoryHeap& heap : memprops.memoryHeaps) {
+        if (heap.flags & vk::MemoryHeapFlagBits::eDeviceLocal) {
+            *free = heap.size;
+            break;
+        }
+    }
+}
+
 // backend registry
 GGML_CALL static ggml_backend_t ggml_backend_reg_vk_init(const char * params, void * user_data) {
     ggml_backend_t vk_backend = ggml_backend_vk_init((int) (intptr_t) user_data);
diff --git a/ggml-vulkan.h b/ggml-vulkan.h
index af661c2d7..c72072e8d 100644
--- a/ggml-vulkan.h
+++ b/ggml-vulkan.h
@@ -19,6 +19,7 @@ GGML_API GGML_CALL bool ggml_backend_is_vk(ggml_backend_t backend);
 GGML_API GGML_CALL int  ggml_backend_vk_get_device_count(void);
 GGML_API GGML_CALL void ggml_backend_vk_get_device_description(int device, char * description, size_t description_size);
 GGML_API GGML_CALL void ggml_backend_vk_get_device_memory(int device, size_t * free, size_t * total);
+GGML_API GGML_CALL void ggml_backend_vk_get_free_device_memory(int device, size_t * free);
 
 GGML_API GGML_CALL ggml_backend_buffer_type_t ggml_backend_vk_buffer_type(size_t dev_num);
 // pinned host buffer for use with the CPU backend for faster copies between CPU and GPU
diff --git a/llama.cpp b/llama.cpp
index 6cad1a553..da72bf0a5 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1659,12 +1659,12 @@ static size_t llama_get_available_device_memory(int device) {
 #elif defined(GGML_USE_SYCL)
     size_t total;
     size_t free;
-    ggml_backend_sycl_get_device_memory(device, &total, &free);
+    ggml_backend_sycl_get_free_device_memory(device, &total, &free);
     return free;
 #elif defined(GGML_USE_VULKAN)
     size_t total;
     size_t free;
-    ggml_backend_vk_get_device_memory(device, &total, &free);
+    ggml_backend_vk_get_free_device_memory(device, &total, &free);
     return free;
 #else
     return 1;