diff --git a/ggml/src/ggml-qnn/backend-ops.cpp b/ggml/src/ggml-qnn/backend-ops.cpp
index d20069874a..c4207e62a3 100644
--- a/ggml/src/ggml-qnn/backend-ops.cpp
+++ b/ggml/src/ggml-qnn/backend-ops.cpp
@@ -573,8 +573,15 @@ bool ggml_qnn_supports_op(ggml_backend_qnn_device_context *ctx, const ggml_tenso
     }
 
     if (op->op == GGML_OP_UNARY) {
-        if (!kQnnUnaryOpsTable[kGgmlUnaryOpStart + ggml_get_unary_op(op)]) {
-            QNN_LOG_DEBUG("unsupported unary op %d", ggml_get_unary_op(op));
+        const auto unary_op = ggml_get_unary_op(op);
+        if (unary_op == GGML_UNARY_OP_GELU && ctx->device == QNN_BACKEND_NPU) {
+            // TODO: fix this when NPU supports GELU
+            QNN_LOG_DEBUG("unsupported unary op GGML_UNARY_OP_GELU for NPU");
+            return false;
+        }
+
+        if (!kQnnUnaryOpsTable[kGgmlUnaryOpStart + unary_op]) {
+            QNN_LOG_DEBUG("unsupported unary op %d", unary_op);
             return false;
         }