diff --git a/tensorflow/core/grappler/optimizers/constant_folding.cc b/tensorflow/core/grappler/optimizers/constant_folding.cc
index 59085473837..b0c3c5b5181 100644
--- a/tensorflow/core/grappler/optimizers/constant_folding.cc
+++ b/tensorflow/core/grappler/optimizers/constant_folding.cc
@@ -901,8 +901,8 @@ DataType GetDataTypeFromNodeOrProps(const NodeDef& node,
 
 // static
 Status ConstantFolding::CreateNodeDef(const string& name,
-                                      const TensorValue& tensor,
-                                      NodeDef* node) {
+                                      const TensorValue& tensor, NodeDef* node,
+                                      size_t original_size) {
   node->set_name(name);
   node->set_op("Const");
 
@@ -980,11 +980,12 @@ Status ConstantFolding::CreateNodeDef(const string& name,
   }
   node->mutable_attr()->insert({"value", attr_tensor});
 
-  if (encoded_size < 10 * 1024 * 1024) {
-    return Status::OK();
+  if (encoded_size > original_size && encoded_size >= 10 * 1024 * 1024) {
+    return errors::InvalidArgument(
+        strings::StrCat("Can't fold ", name, ", its size would be too large (",
+                        encoded_size, " >= ", 10 * 1024 * 1024, " bytes)"));
   }
-  return errors::InvalidArgument(
-      strings::StrCat("Can't fold ", name, ", its size would be too large"));
+  return Status::OK();
 }
 
 Status ConstantFolding::EvaluateNode(const NodeDef& node,
@@ -1010,6 +1011,7 @@ Status ConstantFolding::EvaluateOneFoldable(const NodeDef& node,
     }
   });
 
+  size_t total_inputs_size = 0;
   for (const auto& input : node.input()) {
     const TensorId input_tensor = ParseTensorName(input);
     if (input_tensor.index() < 0) {
@@ -1027,6 +1029,7 @@ Status ConstantFolding::EvaluateOneFoldable(const NodeDef& node,
     Tensor* value = new Tensor(raw_val.dtype(), raw_val.tensor_shape());
     CHECK(value->FromProto(raw_val));
     inputs.emplace_back(value);
+    total_inputs_size += value->TotalBytes();
   }
 
   TF_RETURN_IF_ERROR(EvaluateNode(node, inputs, &output_tensors));
@@ -1041,7 +1044,8 @@ Status ConstantFolding::EvaluateOneFoldable(const NodeDef& node,
       node_name = strings::StrCat(node_name, "-", i);
     }
     if (output_tensors[i].tensor) {
-      Status s = CreateNodeDef(node_name, output_tensors[i], &outputs->at(i));
+      Status s = CreateNodeDef(node_name, output_tensors[i], &outputs->at(i),
+                               total_inputs_size);
       if (!s.ok()) {
         *result_too_large = true;
         return s;
diff --git a/tensorflow/core/grappler/optimizers/constant_folding.h b/tensorflow/core/grappler/optimizers/constant_folding.h
index d6350512f88..99200925cb3 100644
--- a/tensorflow/core/grappler/optimizers/constant_folding.h
+++ b/tensorflow/core/grappler/optimizers/constant_folding.h
@@ -35,8 +35,10 @@ const char kConstantFoldingCtrl[] = "ConstantFoldingCtrl";
 // Constant folding optimization for a graph.
 class ConstantFolding : public GraphOptimizer {
  public:
+  // The size limit will only be considered if the newly created node is greater
+  // than original_size (optional).
   static Status CreateNodeDef(const string& name, const TensorValue& tensor,
-                              NodeDef* node);
+                              NodeDef* node, size_t original_size = 0);
   static string AddControlDependency(const string& input_name, GraphDef* graph,
                                      NodeMap* node_map);