diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 04b6e5285..c4e572c13 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -499,6 +499,8 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_cuda_buffer_type_alloc_buffe
     void * dev_ptr;
     cudaError_t err = cudaMalloc(&dev_ptr, size);
     if (err != cudaSuccess) {
+        // clear the error
+        cudaGetLastError();
         fprintf(stderr, "%s: allocating %.2f MiB on device %d: cudaMalloc failed: %s\n", __func__, size/1024.0/1024.0, buft_ctx->device, cudaGetErrorString(err));
         return nullptr;
     }