diff --git a/onnxruntime/core/optimizer/attention_fusion_helper.h b/onnxruntime/core/optimizer/attention_fusion_helper.h
index 33fff5e1e1a16..a328a6d451a89 100644
--- a/onnxruntime/core/optimizer/attention_fusion_helper.h
+++ b/onnxruntime/core/optimizer/attention_fusion_helper.h
@@ -1447,7 +1447,7 @@ bool FuseGptAttention(Node& layer_norm, Graph& graph, int64_t hidden_size, std::
       return false;
     }
 
-    if (graph_utils::IsSupportedOptypeVersionAndDomain(*k_concat, "Transpose", {1, 13}, kOnnxDomain)) {
+    if (graph_utils::IsSupportedOptypeVersionAndDomain(*k_concat, "Transpose", {1, 13, 21}, kOnnxDomain)) {
       transpose_optimized_pattern = true;
       DEBUG_LOG("Using transpose optimized pattern");
       opt_k_transpose = k_concat;
diff --git a/onnxruntime/core/optimizer/bias_dropout_fusion.cc b/onnxruntime/core/optimizer/bias_dropout_fusion.cc
index 1d4ca9de1d9ff..b93bfedeb1a02 100644
--- a/onnxruntime/core/optimizer/bias_dropout_fusion.cc
+++ b/onnxruntime/core/optimizer/bias_dropout_fusion.cc
@@ -144,7 +144,7 @@ Status BiasDropoutFusion::ApplyImpl(Graph& graph, bool& modified, int graph_leve
     }
 
     const Node& next_node = (*next_node_itr);
-    if ((!graph_utils::IsSupportedOptypeVersionAndDomain(next_node, "Dropout", {12, 13}, kOnnxDomain) &&
+    if ((!graph_utils::IsSupportedOptypeVersionAndDomain(next_node, "Dropout", {12, 13, 22}, kOnnxDomain) &&
          !graph_utils::IsSupportedOptypeVersionAndDomain(next_node, "BitmaskDropout", {1}, kMSDomain)) ||
         next_node.GetExecutionProviderType() != node.GetExecutionProviderType()) {
       continue;
diff --git a/onnxruntime/core/optimizer/compute_optimizer/upstream_reshape.cc b/onnxruntime/core/optimizer/compute_optimizer/upstream_reshape.cc
index 716988e93312c..51a4b38a125b2 100644
--- a/onnxruntime/core/optimizer/compute_optimizer/upstream_reshape.cc
+++ b/onnxruntime/core/optimizer/compute_optimizer/upstream_reshape.cc
@@ -225,7 +225,7 @@ Status UpStreamReshapeGraphTransformer::RemoveOriginalReshapeNode(
 
 std::optional<ReshapeInfo> UpStreamReshapeGraphTransformer::IsSupportedForUpstream(
     Graph& graph, Node& node, const logging::Logger& logger) const {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Reshape", {1, 5, 13, 14}, kOnnxDomain)) {
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Reshape", {1, 5, 13, 14, 19, 21, 23, 24, 25}, kOnnxDomain)) {
     return std::nullopt;
   }
 
diff --git a/onnxruntime/core/optimizer/conv_activation_fusion.cc b/onnxruntime/core/optimizer/conv_activation_fusion.cc
index b7f5af5888be0..44a20428a09e0 100644
--- a/onnxruntime/core/optimizer/conv_activation_fusion.cc
+++ b/onnxruntime/core/optimizer/conv_activation_fusion.cc
@@ -107,7 +107,7 @@ class ConvActivationSelector : public NodeSelector {
       return std::nullopt;
     } else if (node_ep.empty() || node_ep == kCpuExecutionProvider || node_ep == kJsExecutionProvider || node_ep == kWebGpuExecutionProvider) {
       if (!is_supported_non_cuda_ep_activation(*next_node) &&
-          !graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "HardSigmoid", {6})) {
+          !graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "HardSigmoid", {6, 22})) {
         return std::nullopt;
       }
     } else {
@@ -212,7 +212,7 @@ void RegisterConvActivationFusionRules(SelectorActionRegistry& registry) {
   const std::string msDomainConv = SelectorActionRegistry::OpVersionsMapKey("NhwcConv", kMSDomain);
   auto selector = std::make_unique<selectors::ConvActivationSelector>();
 
-  registry.RegisterSelectorAndAction(name, {{"Conv", {1, 11}}, {msInternalNHWCDomainConv, {1, 11}}, {msDomainConv, {1}}},
+  registry.RegisterSelectorAndAction(name, {{"Conv", {1, 11, 22}}, {msInternalNHWCDomainConv, {1, 11, 22}}, {msDomainConv, {1}}},
                                      std::move(selector), std::move(action));
 #else
   registry.RegisterAction(name, std::move(action));
diff --git a/onnxruntime/core/optimizer/conv_add_act_fusion.cc b/onnxruntime/core/optimizer/conv_add_act_fusion.cc
index 6f90eaf07ef4d..45441d20a4112 100644
--- a/onnxruntime/core/optimizer/conv_add_act_fusion.cc
+++ b/onnxruntime/core/optimizer/conv_add_act_fusion.cc
@@ -113,7 +113,7 @@ class ConvAddActivationSelector : public NodeSelector {
         return true;
       }
 
-      if (graph_utils::IsSupportedOptypeVersionAndDomain(activation_node, "HardSigmoid", {6})) {
+      if (graph_utils::IsSupportedOptypeVersionAndDomain(activation_node, "HardSigmoid", {6, 22})) {
         return true;
       }
       return false;
@@ -288,7 +288,7 @@ void RegisterConvAddActivationFusionRules(SelectorActionRegistry& registry) {
   auto action = std::make_unique<actions::FuseConvAddActivationAction>();
   auto selector = std::make_unique<selectors::ConvAddActivationSelector>();
   std::string msDomainNhwcFusedConv = SelectorActionRegistry::OpVersionsMapKey("NhwcFusedConv", kMSDomain);
-  registry.RegisterSelectorAndAction("ConvAddAct", {{"Conv", {1, 11}}, {msDomainNhwcFusedConv, {1, 11}}},
+  registry.RegisterSelectorAndAction("ConvAddAct", {{"Conv", {1, 11, 22}}, {msDomainNhwcFusedConv, {1, 11, 22}}},
                                      std::move(selector), std::move(action));
 }
 
diff --git a/onnxruntime/core/optimizer/conv_add_fusion.cc b/onnxruntime/core/optimizer/conv_add_fusion.cc
index e1fd199bfa943..dd57334666e9f 100644
--- a/onnxruntime/core/optimizer/conv_add_fusion.cc
+++ b/onnxruntime/core/optimizer/conv_add_fusion.cc
@@ -107,7 +107,7 @@ Status ConvAddFusion::Apply(Graph& graph, Node& node, RewriteRuleEffect& modifie
 }
 
 bool ConvAddFusion::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger&) const {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11}) ||
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11, 22}) ||
       node.GetOutputEdgesCount() != 1) {
     return false;
   }
diff --git a/onnxruntime/core/optimizer/conv_bn_fusion.cc b/onnxruntime/core/optimizer/conv_bn_fusion.cc
index 4c493f45a2b61..dff11f2276097 100644
--- a/onnxruntime/core/optimizer/conv_bn_fusion.cc
+++ b/onnxruntime/core/optimizer/conv_bn_fusion.cc
@@ -145,7 +145,7 @@ Status ConvBNFusion::Apply(Graph& graph, Node& node, RewriteRuleEffect& rule_eff
 }
 
 bool ConvBNFusion::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger&) const {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11}) ||
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11, 22}) ||
       node.GetOutputEdgesCount() != 1) {
     return false;
   }
diff --git a/onnxruntime/core/optimizer/conv_mul_fusion.cc b/onnxruntime/core/optimizer/conv_mul_fusion.cc
index 9563415ad56b6..4dd10edbbfb89 100644
--- a/onnxruntime/core/optimizer/conv_mul_fusion.cc
+++ b/onnxruntime/core/optimizer/conv_mul_fusion.cc
@@ -113,7 +113,7 @@ Status ConvMulFusion::Apply(Graph& graph, Node& node, RewriteRuleEffect& rule_ef
 }
 
 bool ConvMulFusion::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger&) const {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11}) ||
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11, 22}) ||
       node.GetOutputEdgesCount() != 1) {
     return false;
   }
diff --git a/onnxruntime/core/optimizer/dropout_elimination.cc b/onnxruntime/core/optimizer/dropout_elimination.cc
index d989c4dd80532..b76f6289d5fc6 100644
--- a/onnxruntime/core/optimizer/dropout_elimination.cc
+++ b/onnxruntime/core/optimizer/dropout_elimination.cc
@@ -22,7 +22,7 @@ Status EliminateDropout::Apply(Graph& graph, Node& node, RewriteRuleEffect& rule
 bool EliminateDropout::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger& logger) const {
   // We currently support elimination for Dropout operator v1, v6, v7, v10 and v12.
   // REVIEW(mzs): v10 implementation does not exist.
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Dropout", {1, 6, 7, 10, 12, 13})) {
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Dropout", {1, 6, 7, 10, 12, 13, 22})) {
     return false;
   }
 
@@ -32,7 +32,7 @@ bool EliminateDropout::SatisfyCondition(const Graph& graph, const Node& node, co
   //    2. ratio input is not a graph input, so it cannot be overridden
 
   // support opset 12 and above for ort training
-  if (graph_utils::MatchesOpSinceVersion(node, {12, 13}) && node.InputDefs().size() > 1) {
+  if (graph_utils::MatchesOpSinceVersion(node, {12, 13, 22}) && node.InputDefs().size() > 1) {
     if (graph_utils::IsGraphInput(graph, node.InputDefs()[1])) {
       return false;
     }
diff --git a/onnxruntime/core/optimizer/fast_gelu_fusion.cc b/onnxruntime/core/optimizer/fast_gelu_fusion.cc
index a38a0fc06fbb2..3d8e04b243197 100644
--- a/onnxruntime/core/optimizer/fast_gelu_fusion.cc
+++ b/onnxruntime/core/optimizer/fast_gelu_fusion.cc
@@ -151,7 +151,7 @@ MatchResult FastGeluFusion::CheckSecondFormula(Graph& graph, Node& pow1_node,
   if (p_cast1_node != nullptr) {
     Node& cast1_node = *graph.GetNode(p_cast1_node->Index());
     // this is fused Cast node, so expect 2 output edges
-    if (!(graph_utils::IsSupportedOptypeVersionAndDomain(cast1_node, "Cast", {9, 13, 19}) &&
+    if (!(graph_utils::IsSupportedOptypeVersionAndDomain(cast1_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
           CheckNode(graph, cast1_node, pow1_node.GetExecutionProviderType(), false)) ||
         cast1_node.GetOutputEdgesCount() != 2) {
       return matchResult;
@@ -262,7 +262,7 @@ Status FastGeluFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level,
         if (p_cast3_node == nullptr) continue;
 
         Node& cast3_node = *graph.GetNode(p_cast3_node->Index());
-        if (!(graph_utils::IsSupportedOptypeVersionAndDomain(cast3_node, "Cast", {9, 13, 19}) &&
+        if (!(graph_utils::IsSupportedOptypeVersionAndDomain(cast3_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
               CheckNode(graph, cast3_node, node.GetExecutionProviderType(), true))) {
           continue;
         }
diff --git a/onnxruntime/core/optimizer/gemm_activation_fusion.cc b/onnxruntime/core/optimizer/gemm_activation_fusion.cc
index 50be2cbd48f7b..7465607ac7124 100644
--- a/onnxruntime/core/optimizer/gemm_activation_fusion.cc
+++ b/onnxruntime/core/optimizer/gemm_activation_fusion.cc
@@ -21,20 +21,20 @@ bool IsSupportedOptypeVersionAndDomain(const Node& node, const std::string& op_t
 // If the op has multiple versions, here we require it must have a single implementation that can work across all the
 // versions. Because in the fusion, we discarded the op version information.
 bool IsFusableActivation(const Node& node) {
-  return IsSupportedOptypeVersionAndDomain(node, "Elu", {6}, kOnnxDomain) ||
-         IsSupportedOptypeVersionAndDomain(node, "HardSigmoid", {6}, kOnnxDomain) ||
-         IsSupportedOptypeVersionAndDomain(node, "LeakyRelu", {6}, kOnnxDomain) ||
+  return IsSupportedOptypeVersionAndDomain(node, "Elu", {6, 22}, kOnnxDomain) ||
+         IsSupportedOptypeVersionAndDomain(node, "HardSigmoid", {6, 22}, kOnnxDomain) ||
+         IsSupportedOptypeVersionAndDomain(node, "LeakyRelu", {6, 16}, kOnnxDomain) ||
          IsSupportedOptypeVersionAndDomain(node, "Relu", {6, 13, 14}, kOnnxDomain) ||
-         IsSupportedOptypeVersionAndDomain(node, "Selu", {6}, kOnnxDomain) ||
+         IsSupportedOptypeVersionAndDomain(node, "Selu", {6, 22}, kOnnxDomain) ||
          IsSupportedOptypeVersionAndDomain(node, "Sigmoid", {6, 13}, kOnnxDomain) ||
-         IsSupportedOptypeVersionAndDomain(node, "Softplus", {1}, kOnnxDomain) ||
-         IsSupportedOptypeVersionAndDomain(node, "Softsign", {1}, kOnnxDomain) ||
+         IsSupportedOptypeVersionAndDomain(node, "Softplus", {1, 22}, kOnnxDomain) ||
+         IsSupportedOptypeVersionAndDomain(node, "Softsign", {1, 22}, kOnnxDomain) ||
          IsSupportedOptypeVersionAndDomain(node, "Tanh", {6, 13}, kOnnxDomain) ||
 #ifndef DISABLE_CONTRIB_OPS
          IsSupportedOptypeVersionAndDomain(node, "ScaledTanh", {1}, kOnnxDomain) ||
          IsSupportedOptypeVersionAndDomain(node, "ParametricSoftplus", {1}, kOnnxDomain) ||
 #endif
-         IsSupportedOptypeVersionAndDomain(node, "ThresholdedRelu", {1, 10}, kOnnxDomain);
+         IsSupportedOptypeVersionAndDomain(node, "ThresholdedRelu", {1, 10, 22}, kOnnxDomain);
 }
 }  // namespace
 
diff --git a/onnxruntime/core/optimizer/gemm_transpose_fusion.cc b/onnxruntime/core/optimizer/gemm_transpose_fusion.cc
index d1f862460dae7..da454b67aecf4 100644
--- a/onnxruntime/core/optimizer/gemm_transpose_fusion.cc
+++ b/onnxruntime/core/optimizer/gemm_transpose_fusion.cc
@@ -104,7 +104,7 @@ bool GemmTransposeFusion::SatisfyCondition(const Graph& graph, const Node& node,
 
   // Fusion can be applied if there is a transpose at either of the inputs
   for (auto node_it = node.InputNodesBegin(); node_it != node.InputNodesEnd(); ++node_it) {
-    if (graph_utils::IsSupportedOptypeVersionAndDomain(*node_it, "Transpose", {1, 13}) &&
+    if (graph_utils::IsSupportedOptypeVersionAndDomain(*node_it, "Transpose", {1, 13, 21, 23, 24, 25}) &&
         !graph.NodeProducesGraphOutput(*node_it) &&
         // Make sure the two nodes do not span execution providers.
         node_it->GetExecutionProviderType() == node.GetExecutionProviderType()) {
@@ -128,7 +128,7 @@ bool GemmTransposeFusion::SatisfyCondition(const Graph& graph, const Node& node,
 
   const auto next_node_it = node.OutputNodesBegin();
   if (next_node_it != node.OutputNodesEnd() &&
-      graph_utils::IsSupportedOptypeVersionAndDomain(*next_node_it, "Transpose", {1, 13}) &&
+      graph_utils::IsSupportedOptypeVersionAndDomain(*next_node_it, "Transpose", {1, 13, 21, 23, 24, 25}) &&
       next_node_it->GetInputEdgesCount() == 1 &&
       // Make sure the two nodes do not span execution providers.
       next_node_it->GetExecutionProviderType() == node.GetExecutionProviderType()) {
diff --git a/onnxruntime/core/optimizer/isinf_reducesum_fusion.cc b/onnxruntime/core/optimizer/isinf_reducesum_fusion.cc
index 7d249ea715e8d..3f963e86acb0d 100644
--- a/onnxruntime/core/optimizer/isinf_reducesum_fusion.cc
+++ b/onnxruntime/core/optimizer/isinf_reducesum_fusion.cc
@@ -33,7 +33,7 @@ Status IsInfReduceSumFusion::ApplyImpl(Graph& graph, bool& modified, int graph_l
 
     ORT_RETURN_IF_ERROR(Recurse(isinf_node, modified, graph_level, logger));
 
-    if (!graph_utils::IsSupportedOptypeVersionAndDomain(isinf_node, "IsInf", {10}) ||
+    if (!graph_utils::IsSupportedOptypeVersionAndDomain(isinf_node, "IsInf", {10, 20}) ||
         isinf_node.GetOutputEdgesCount() != 1 ||
         graph.NodeProducesGraphOutput(isinf_node)) {
       continue;
@@ -45,7 +45,7 @@ Status IsInfReduceSumFusion::ApplyImpl(Graph& graph, bool& modified, int graph_l
     // This Cast can be skipped as we are replacing the subgraph with IsAllFinite, which supports FP16
     auto cast1_node_iter = isinf_node.InputNodesBegin();
     if (cast1_node_iter != isinf_node.InputNodesEnd() &&
-        graph_utils::IsSupportedOptypeVersionAndDomain(*cast1_node_iter, "Cast", {9, 13, 19}) &&
+        graph_utils::IsSupportedOptypeVersionAndDomain(*cast1_node_iter, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
         cast1_node_iter->GetOutputEdgesCount() == 1) {
       // check input type of cast node
       Node& cast1_node = *graph.GetNode(cast1_node_iter->Index());
@@ -65,7 +65,7 @@ Status IsInfReduceSumFusion::ApplyImpl(Graph& graph, bool& modified, int graph_l
     }
 
     Node& cast2_node = *graph.GetNode(cast2_node_itr->Index());
-    if (!graph_utils::IsSupportedOptypeVersionAndDomain(cast2_node, "Cast", {9, 13, 19}) ||
+    if (!graph_utils::IsSupportedOptypeVersionAndDomain(cast2_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) ||
         cast2_node.GetOutputEdgesCount() != 1 ||
         graph.NodeProducesGraphOutput(cast2_node)) {
       continue;
diff --git a/onnxruntime/core/optimizer/layer_norm_fusion.cc b/onnxruntime/core/optimizer/layer_norm_fusion.cc
index 8a7f83e871768..3ade3864255ea 100644
--- a/onnxruntime/core/optimizer/layer_norm_fusion.cc
+++ b/onnxruntime/core/optimizer/layer_norm_fusion.cc
@@ -241,7 +241,7 @@ Status LayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level,
     if (p_reduce_mean_input_node) {
       Node& reduce_mean_input_node = *graph.GetNode(p_reduce_mean_input_node->Index());
       // If input to the 1st ReduceMean is a Cast, and the Cast has same consumer count as subCnt + 1
-      if (graph_utils::IsSupportedOptypeVersionAndDomain(reduce_mean_input_node, "Cast", {9, 13, 19}) &&
+      if (graph_utils::IsSupportedOptypeVersionAndDomain(reduce_mean_input_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
           reduce_mean_input_node.GetExecutionProviderType() == reduce_mean_node.GetExecutionProviderType() &&
           optimizer_utils::CheckOutputEdges(graph, reduce_mean_input_node, static_cast<size_t>(subCnt) + 1)) {
         nodes_to_remove.insert(nodes_to_remove.begin(), reduce_mean_input_node);
@@ -254,7 +254,7 @@ Status LayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level,
     const Node* p_cast1 = nullptr;
     if (!p_sub_node_dup && sub_node.GetOutputEdgesCount() == 1) {
       Node& cast_node = *graph.GetNode(sub_node.OutputNodesBegin()->Index());
-      if (graph_utils::IsSupportedOptypeVersionAndDomain(cast_node, "Cast", {9, 13, 19}) &&
+      if (graph_utils::IsSupportedOptypeVersionAndDomain(cast_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
           cast_node.GetExecutionProviderType() == reduce_mean_node.GetExecutionProviderType() &&
           optimizer_utils::CheckOutputEdges(graph, cast_node, 2u) && IsSupportedDataType(cast_node)) {
         p_cast1 = &cast_node;
@@ -353,7 +353,7 @@ Status LayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level,
     const Node* p_cast2 = graph_utils::FirstParentByType(pow_node, "Cast");
     if (p_cast2 != nullptr && p_cast2 != p_cast1) {
       Node& cast_node = *graph.GetNode(p_cast2->Index());
-      if (!graph_utils::IsSupportedOptypeVersionAndDomain(cast_node, "Cast", {9, 13, 19}) ||
+      if (!graph_utils::IsSupportedOptypeVersionAndDomain(cast_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) ||
           cast_node.GetExecutionProviderType() != reduce_mean_node.GetExecutionProviderType() ||
           !optimizer_utils::CheckOutputEdges(graph, cast_node, 1)) {
         continue;
@@ -371,7 +371,7 @@ Status LayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level,
     // can be removed. This is one possible place a Cast Op can exist, that is between Div and Mul nodes.
     // div --> mul or div --> cast --> mul
     Node* next_node = graph.GetNode(div_node.OutputNodesBegin()->Index());
-    if (graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Cast", {9, 13, 19}) &&
+    if (graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
         optimizer_utils::CheckOutputEdges(graph, *next_node, 1)) {
       nodes_to_remove.push_back(*next_node);
       next_node = graph.GetNode(next_node->OutputNodesBegin()->Index());
@@ -637,7 +637,7 @@ Status SimplifiedLayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int gr
     if (is_gpu_ep && p_pow_input_node) {
       Node& pow_input_node = *graph.GetNode(p_pow_input_node->Index());
       // If input to Pow is a Cast, and the Cast has 2 consumers only (Pow, Div)
-      if (graph_utils::IsSupportedOptypeVersionAndDomain(pow_input_node, "Cast", {9, 13, 19}) &&
+      if (graph_utils::IsSupportedOptypeVersionAndDomain(pow_input_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
           pow_input_node.GetExecutionProviderType() == pow_node.GetExecutionProviderType() &&
           optimizer_utils::CheckOutputEdges(graph, pow_input_node, 2)) {
         nodes_to_remove.insert(nodes_to_remove.begin(), pow_input_node);
@@ -647,7 +647,7 @@ Status SimplifiedLayerNormFusion::ApplyImpl(Graph& graph, bool& modified, int gr
 
     // div --> mul or div --> cast --> mul
     Node* next_node = graph.GetNode(div_node.OutputNodesBegin()->Index());
-    if (graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Cast", {9, 13, 19}) &&
+    if (graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Cast", {9, 13, 19, 21, 23, 24, 25}) &&
         optimizer_utils::CheckOutputEdges(graph, *next_node, 1)) {
       if (!is_gpu_ep) continue;
       nodes_to_remove.push_back(*next_node);
diff --git a/onnxruntime/core/optimizer/matmul_bn_fusion.cc b/onnxruntime/core/optimizer/matmul_bn_fusion.cc
index 63a15ad630ce9..871571ea64881 100644
--- a/onnxruntime/core/optimizer/matmul_bn_fusion.cc
+++ b/onnxruntime/core/optimizer/matmul_bn_fusion.cc
@@ -10,8 +10,8 @@ namespace onnxruntime {
 
 namespace {
 const std::vector<std::pair<std::string, InlinedVector<ONNX_NAMESPACE::OperatorSetVersion>>> ignorable_nodes{
-    {"Reshape", {1, 5, 13, 14, 19}},
-    {"Transpose", {1, 13}}};
+    {"Reshape", {1, 5, 13, 14, 19, 21, 23, 24, 25}},
+    {"Transpose", {1, 13, 21, 23, 24, 25}}};
 const std::pair<std::string, InlinedVector<ONNX_NAMESPACE::OperatorSetVersion>> dest = {"BatchNormalization", {1, 6, 7, 9, 14, 15}};
 }  // namespace
 
@@ -244,4 +244,4 @@ Status MatmulBNFusion::Apply(Graph& graph, Node& matmul_node, RewriteRuleEffect&
   rule_effect = RewriteRuleEffect::kRemovedCurrentNode;
   return Status::OK();
 }
-}  // namespace onnxruntime
\ No newline at end of file
+}  // namespace onnxruntime
diff --git a/onnxruntime/core/optimizer/nchwc_transformer.cc b/onnxruntime/core/optimizer/nchwc_transformer.cc
index bff9d2990118a..e8c3bf24a612f 100644
--- a/onnxruntime/core/optimizer/nchwc_transformer.cc
+++ b/onnxruntime/core/optimizer/nchwc_transformer.cc
@@ -1149,15 +1149,15 @@ void NchwcTransformerImpl::TrackTransposeFromNhwc(Node& node) {
 }
 
 void NchwcTransformerImpl::Transform(Node& node) {
-  if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13})) {
+  if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13, 21, 23, 24, 25})) {
     TrackTransposeFromNhwc(node);
   }
 
-  if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11}) ||
+  if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Conv", {1, 11, 22}) ||
       graph_utils::IsSupportedOptypeVersionAndDomain(node, "FusedConv", {1}, kMSDomain)) {
     TransformConv(node);
-  } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "MaxPool", {1, 8, 10, 11, 12}) ||
-             graph_utils::IsSupportedOptypeVersionAndDomain(node, "AveragePool", {1, 7, 10, 11})) {
+  } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "MaxPool", {1, 8, 10, 11, 12, 22}) ||
+             graph_utils::IsSupportedOptypeVersionAndDomain(node, "AveragePool", {1, 7, 10, 11, 19, 22})) {
     TransformPool(node);
   } else if (node.GetInputEdgesCount() == 0 && node.InputDefs().size() != 0) {
     // The following transforms only run when the input edge count has already
@@ -1176,15 +1176,15 @@ void NchwcTransformerImpl::Transform(Node& node) {
                graph_utils::IsSupportedOptypeVersionAndDomain(node, "Sigmoid", {6, 13}) ||
                graph_utils::IsSupportedOptypeVersionAndDomain(node, "Tanh", {6, 13})) {
       TransformActivation(node);
-    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "BatchNormalization", {7, 9, 14})) {
+    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "BatchNormalization", {7, 9, 14, 15})) {
       TransformBatchNormalization(node);
-    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13})) {
+    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13, 21, 23, 24, 25})) {
       TransformTransposeToNhwc(node);
     } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "Upsample", {9, 13}) ||
-               graph_utils::IsSupportedOptypeVersionAndDomain(node, "Resize", {10, 11, 13})) {
+               graph_utils::IsSupportedOptypeVersionAndDomain(node, "Resize", {10, 11, 13, 18, 19})) {
       TransformResize(node);
-    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "GlobalMaxPool", {1}) ||
-               graph_utils::IsSupportedOptypeVersionAndDomain(node, "GlobalAveragePool", {1})) {
+    } else if (graph_utils::IsSupportedOptypeVersionAndDomain(node, "GlobalMaxPool", {1, 22}) ||
+               graph_utils::IsSupportedOptypeVersionAndDomain(node, "GlobalAveragePool", {1, 22})) {
       // Convert these pooling types only if the input is already in NCHWc format.
       TransformPool(node);
     }
diff --git a/onnxruntime/core/optimizer/not_where_fusion.cc b/onnxruntime/core/optimizer/not_where_fusion.cc
index 862b192a10cce..618374f8e5347 100644
--- a/onnxruntime/core/optimizer/not_where_fusion.cc
+++ b/onnxruntime/core/optimizer/not_where_fusion.cc
@@ -39,7 +39,7 @@ Condition -> Where ->
           v0----|
  */
 bool NotWhereFusion::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger& logger) const {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Where", {9})) {
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Where", {9, 16})) {
     return false;
   }
 
@@ -54,7 +54,7 @@ bool NotWhereFusion::SatisfyCondition(const Graph& graph, const Node& node, cons
   if (p_not_node->GetOutputEdgesCount() > 1) {
     // all consumers of not must be where
     for (auto it = p_not_node->OutputNodesBegin(); it != p_not_node->OutputNodesEnd(); ++it) {
-      if (!graph_utils::IsSupportedOptypeVersionAndDomain(*it, "Where", {9})) {
+      if (!graph_utils::IsSupportedOptypeVersionAndDomain(*it, "Where", {9, 16})) {
         return false;
       }
     }
diff --git a/onnxruntime/core/optimizer/pad_fusion.cc b/onnxruntime/core/optimizer/pad_fusion.cc
index d0b6d42fd46c9..b364b770a425c 100644
--- a/onnxruntime/core/optimizer/pad_fusion.cc
+++ b/onnxruntime/core/optimizer/pad_fusion.cc
@@ -9,9 +9,9 @@
 namespace onnxruntime {
 
 bool VerifyNotCastChild(const Node& child_node) {
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "Conv", {1, 11}) &&
-      !graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "AveragePool", {7, 10, 11, 19}) &&
-      !graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "MaxPool", {1, 8, 10, 11, 12})) {
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "Conv", {1, 11, 22}) &&
+      !graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "AveragePool", {7, 10, 11, 19, 22}) &&
+      !graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "MaxPool", {1, 8, 10, 11, 12, 22})) {
     return false;
   }
 
@@ -90,7 +90,7 @@ void UpdatePaddingAttribute(Node& child_node, const std::vector<int64_t>& pads_v
  */
 bool PadFusion::SatisfyCondition(const Graph& graph, const Node& node, const logging::Logger&) const {
   // if Pad has input axis, don't fuse it.
-  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Pad", {1, 2, 11, 13, 18, 19}) ||
+  if (!graph_utils::IsSupportedOptypeVersionAndDomain(node, "Pad", {1, 2, 11, 13, 18, 19, 21, 23, 24, 25}) ||
       node.GetOutputEdgesCount() != 1 ||
       node.InputDefs().size() > 3) {
     return false;
@@ -130,7 +130,7 @@ bool PadFusion::SatisfyCondition(const Graph& graph, const Node& node, const log
   }
 
   const Node& child_node = *node.OutputNodesBegin();
-  if (graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "Cast", {1, 6, 9, 13})) {
+  if (graph_utils::IsSupportedOptypeVersionAndDomain(child_node, "Cast", {1, 6, 9, 13, 19, 21, 23, 24, 25})) {
     if (child_node.GetOutputEdgesCount() != 1) {
       return false;
     }
diff --git a/onnxruntime/core/optimizer/pre_shape_node_elimination.cc b/onnxruntime/core/optimizer/pre_shape_node_elimination.cc
index 8f50ef7c09c95..f64ee1d4ec8f5 100644
--- a/onnxruntime/core/optimizer/pre_shape_node_elimination.cc
+++ b/onnxruntime/core/optimizer/pre_shape_node_elimination.cc
@@ -48,7 +48,7 @@ bool PreShapeNodeElimination::SatisfyCondition(const Graph& graph, const Node& n
 
   for (const Node* next_node : output_nodes) {
     // Check if the next node is not of type "Shape"
-    if (!next_node || !graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Shape", {13, 15, 19}, kOnnxDomain)) {
+    if (!next_node || !graph_utils::IsSupportedOptypeVersionAndDomain(*next_node, "Shape", {13, 15, 19, 21, 23, 24, 25}, kOnnxDomain)) {
       return false;
     }
   }
diff --git a/onnxruntime/core/optimizer/qdq_transformer/qdq_propagation.cc b/onnxruntime/core/optimizer/qdq_transformer/qdq_propagation.cc
index 7b518947138a5..b8252bc7a75b4 100644
--- a/onnxruntime/core/optimizer/qdq_transformer/qdq_propagation.cc
+++ b/onnxruntime/core/optimizer/qdq_transformer/qdq_propagation.cc
@@ -21,11 +21,11 @@ using onnxruntime::graph_utils::ExtendedGraphEdge;
 namespace onnxruntime {
 namespace {
 bool CanNodePropagate(const Node& node) {
-  return graph_utils::IsSupportedOptypeVersionAndDomain(node, "MaxPool", {12}) ||
-         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Reshape", {5, 13, 14, 19, 21}) ||
-         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13, 21}) ||
-         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Squeeze", {1, 11, 13, 21}) ||
-         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Unsqueeze", {1, 11, 13, 21}) ||
+  return graph_utils::IsSupportedOptypeVersionAndDomain(node, "MaxPool", {12, 22}) ||
+         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Reshape", {5, 13, 14, 19, 21, 23, 24, 25}) ||
+         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Transpose", {1, 13, 21, 23, 24, 25}) ||
+         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Squeeze", {1, 11, 13, 21, 23, 24, 25}) ||
+         graph_utils::IsSupportedOptypeVersionAndDomain(node, "Unsqueeze", {1, 11, 13, 21, 23, 24, 25}) ||
          graph_utils::IsSupportedOptypeVersionAndDomain(node, "Slice", {1, 10, 11, 13});
 }
 
diff --git a/onnxruntime/core/optimizer/reshape_fusion.cc b/onnxruntime/core/optimizer/reshape_fusion.cc
index daab9bba278aa..6a2b4295093d8 100644
--- a/onnxruntime/core/optimizer/reshape_fusion.cc
+++ b/onnxruntime/core/optimizer/reshape_fusion.cc
@@ -34,7 +34,7 @@ Status ReshapeFusion::ApplyImpl(Graph& graph, bool& modified, int graph_level, c
     Node& reshape = *p_reshape;
     ORT_RETURN_IF_ERROR(Recurse(reshape, modified, graph_level, logger));
 
-    if (!graph_utils::IsSupportedOptypeVersionAndDomain(reshape, "Reshape", {5, 13, 14}) ||
+    if (!graph_utils::IsSupportedOptypeVersionAndDomain(reshape, "Reshape", {5, 13, 14, 19, 21, 23, 24, 25}) ||
         !graph_utils::IsSupportedProvider(reshape, GetCompatibleExecutionProviders())) {
       continue;
     }
@@ -92,8 +92,8 @@ static bool Match_Linear_Subgraph_1(Graph& graph, const Node& concat, const Node
   const Node& reshape = *reshape_itr;
 
   std::vector<graph_utils::EdgeEndToMatch> linear_path{
-      {0, 0, "Add", {7}, kOnnxDomain},
-      {0, 0, "MatMul", {1, 9}, kOnnxDomain}};
+      {0, 0, "Add", {7, 13, 14}, kOnnxDomain},
+      {0, 0, "MatMul", {1, 9, 13}, kOnnxDomain}};
   std::vector<const Node::EdgeEnd*> edges;
   if (!graph_utils::FindPath(reshape, true, linear_path, edges, logger)) {
     return false;
@@ -158,9 +158,9 @@ bool ReshapeFusion::Match_One_Element_Output_Subgraph_1(Graph& graph, const Node
                                                         int index, gsl::span<const int64_t> shape_value, bool checkOneElementOnly,
                                                         const logging::Logger& logger) {
   std::vector<graph_utils::EdgeEndToMatch> parent_path{
-      {0, index, "Unsqueeze", {1, 11, 13}, kOnnxDomain},
+      {0, index, "Unsqueeze", {1, 11, 13, 21, 23, 24, 25}, kOnnxDomain},
       {0, 0, "Gather", {1, 11, 13}, kOnnxDomain},
-      {0, 0, "Shape", {1, 13, 15}, kOnnxDomain}};
+      {0, 0, "Shape", {1, 13, 15, 19, 21, 23, 24, 25}, kOnnxDomain}};
   std::vector<const Node::EdgeEnd*> edges;
   if (graph_utils::FindPath(concat, true, parent_path, edges, logger)) {
     const Node& unsqueeze = edges[0]->GetNode();
@@ -204,9 +204,9 @@ bool ReshapeFusion::Match_One_Element_Output_Subgraph_1(Graph& graph, const Node
 bool ReshapeFusion::Match_One_Element_Output_Subgraph_2(Graph& graph, const NodeArg& root_input, const Node& cur_node,
                                                         int index, const logging::Logger& logger) {
   std::vector<graph_utils::EdgeEndToMatch> parent_path{
-      {0, index, "Squeeze", {1, 11, 13}, kOnnxDomain},
+      {0, index, "Squeeze", {1, 11, 13, 21, 23, 24, 25}, kOnnxDomain},
       {0, 0, "Slice", {1, 11, 13}, kOnnxDomain},
-      {0, 0, "Shape", {1, 13}, kOnnxDomain}};
+      {0, 0, "Shape", {1, 13, 15, 19, 21, 23, 24, 25}, kOnnxDomain}};
   std::vector<const Node::EdgeEnd*> edges;
   if (graph_utils::FindPath(cur_node, true, parent_path, edges, logger)) {
     const Node& slice = edges[1]->GetNode();
@@ -282,15 +282,15 @@ bool ReshapeFusion::Is_One_Element_Output_Subgraph(Graph& graph, const NodeArg&
   }
 
   std::vector<graph_utils::EdgeEndToMatch> div_path{
-      {0, index, "Unsqueeze", {1, 11, 13}, kOnnxDomain},
+      {0, index, "Unsqueeze", {1, 11, 13, 21, 23, 24, 25}, kOnnxDomain},
       {0, 0, "Div", {7, 13, 14}, kOnnxDomain}};
 
   std::vector<graph_utils::EdgeEndToMatch> mul_path{
-      {0, index, "Unsqueeze", {1, 11, 13}, kOnnxDomain},
+      {0, index, "Unsqueeze", {1, 11, 13, 21, 23, 24, 25}, kOnnxDomain},
       {0, 0, "Mul", {7, 13, 14}, kOnnxDomain}};
 
   std::vector<graph_utils::EdgeEndToMatch> unsqueeze_path{
-      {0, index, "Unsqueeze", {1, 11, 13}, kOnnxDomain}};
+      {0, index, "Unsqueeze", {1, 11, 13, 21, 23, 24, 25}, kOnnxDomain}};
 
   std::vector<const Node::EdgeEnd*> edges;
   if (graph_utils::FindPath(concat, true, div_path, edges, logger) ||
diff --git a/onnxruntime/test/optimizer/conv_add_act_test.cc b/onnxruntime/test/optimizer/conv_add_act_test.cc
index f61f9b29d9cce..bb409a2bbb82e 100644
--- a/onnxruntime/test/optimizer/conv_add_act_test.cc
+++ b/onnxruntime/test/optimizer/conv_add_act_test.cc
@@ -36,8 +36,8 @@ void TestConvPath(const std::vector<int64_t>& input_shape, const std::vector<int
   TransformerTester(build_test_case,
                     check_graph,
                     TransformerLevel::Default,
-                    TransformerLevel::Level3, 12, 0.0001, 0.000001,
-                    0, {}, disabled_optimizers);
+                    TransformerLevel::Level3, {12, 22}, 0.0001, 0.000001,
+                    nullptr, {}, disabled_optimizers);
 }
 
 TEST(ConvAddActivationFusionTests, ConvExpandThenGemm) {
diff --git a/onnxruntime/test/optimizer/graph_transform_test.cc b/onnxruntime/test/optimizer/graph_transform_test.cc
index 70e84733fa869..d7780da36626c 100644
--- a/onnxruntime/test/optimizer/graph_transform_test.cc
+++ b/onnxruntime/test/optimizer/graph_transform_test.cc
@@ -7059,7 +7059,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_ShareIntTypedInitializer) {
       }
     };
 
-    const std::vector<int> opsets{12, 13, 14};  // Clip support int64_t since opset 12
+    const std::vector<int> opsets{12, 13, 14, 23};  // Clip support int64_t since opset 12
     for (auto& opset_version : opsets) {
       std::unique_ptr<GraphTransformer> transformer = std::make_unique<ConstantSharing>();
       ASSERT_STATUS_OK(TestGraphTransformer(build_test_case, opset_version, *logger_, std::move(transformer), TransformerLevel::Level1,
@@ -7156,7 +7156,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_ShareFloatOrHalfTypedInitialize
     return Status::OK();
   };
 
-  const std::vector<int> opsets{12, 13, 14};  // Clip support int64_t since opset 12
+  const std::vector<int> opsets{12, 13, 14, 23};  // Clip support int64_t since opset 12
 
   // Float data type tests.
   auto build_test_case_float = [&](ModelTestBuilder& builder) {
@@ -7280,7 +7280,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_Share2DFloatOrHalfTypedInitiali
     return Status::OK();
   };
 
-  const std::vector<int> opsets{12, 13, 14};  // Clip support int64_t since opset 12
+  const std::vector<int> opsets{12, 13, 14, 23};  // Clip support int64_t since opset 12
 
   // Float data type tests.
   auto build_test_case_float = [&](ModelTestBuilder& builder) {
@@ -7386,7 +7386,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_ShareFloatAndHalfTypedInitializ
     return Status::OK();
   };
 
-  const std::vector<int> opsets{12, 13, 14};
+  const std::vector<int> opsets{12, 13, 14, 23};
 
   auto build_test_case_float = [&](ModelTestBuilder& builder) {
     auto* input0_arg = builder.MakeInput<float>({{1, 1, 256, 256}});
@@ -7525,7 +7525,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_Share2DFloatAndHalfTypedInitial
     return Status::OK();
   };
 
-  const std::vector<int> opsets{12, 13, 14};
+  const std::vector<int> opsets{12, 13, 14, 23};
 
   std::vector<float> values{0.0f, 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f};
   std::vector<MLFloat16> values_float16;
@@ -7651,7 +7651,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_ShareIntMaxOrFloatInfinityIniti
     return Status::OK();
   };
 
-  const std::vector<int> opsets{12, 13, 14};
+  const std::vector<int> opsets{12, 13, 14, 23};
 
   // Float data type tests.
   auto build_test_case_float = [&](ModelTestBuilder& builder) {
diff --git a/onnxruntime/test/optimizer/nchwc_optimizer_test.cc b/onnxruntime/test/optimizer/nchwc_optimizer_test.cc
index fc0ba86c7f1f6..7dfa5c7812f6e 100644
--- a/onnxruntime/test/optimizer/nchwc_optimizer_test.cc
+++ b/onnxruntime/test/optimizer/nchwc_optimizer_test.cc
@@ -553,7 +553,7 @@ TEST(NchwcOptimizerTests, ConvAddFusion) {
   // Verify that Add or Sum can be fused into a preceding NCHWc Conv node,
   // with an optional Relu node following.
   std::vector<std::string> op_types{"Add", "Sum"};
-  static const int opset_versions[] = {7, 10, 11, 12};
+  static const int opset_versions[] = {7, 10, 11, 12, 14, 22};
   for (auto& op_type : op_types) {
     for (auto opset_version : opset_versions) {
       test_case(op_type, opset_version, false);
@@ -1271,7 +1271,7 @@ TEST(NchwcOptimizerTests, UpsampleNearest) {
 
   // Verify that upsample nodes can be converted to the NCHWc format for
   // various versions of the operator.
-  static const int opset_versions[] = {9, 10, 11, 13};
+  static const int opset_versions[] = {9, 10, 11, 13, 18, 19};
   for (auto opset_version : opset_versions) {
     test_case(opset_version, 1.f, 1.f, false);
     test_case(opset_version, 2.f, 2.f, false);
@@ -1327,7 +1327,7 @@ TEST(NchwcOptimizerTests, UpsampleLinear) {
   // various versions of the operator.
   std::vector<std::string> transformation_modes{"asymmetric", "align_corners", "half_pixel"};
   for (auto& transformation_mode : transformation_modes) {
-    static const int opset_versions[] = {9, 10, 11, 13};
+    static const int opset_versions[] = {9, 10, 11, 13, 18, 19};
     for (auto opset_version : opset_versions) {
       // Older versions of the operator do not support transformation modes.
       if (opset_version < 11 && transformation_mode == "asymmetric") {
diff --git a/onnxruntime/test/optimizer/qdq_transformer_test.cc b/onnxruntime/test/optimizer/qdq_transformer_test.cc
index 7fe68a38d23a0..37da2d4247e34 100644
--- a/onnxruntime/test/optimizer/qdq_transformer_test.cc
+++ b/onnxruntime/test/optimizer/qdq_transformer_test.cc
@@ -3584,6 +3584,11 @@ TEST(QDQTransformerTests, QDQPropagation_QBackward) {
                       check_graph,
                       TransformerLevel::Default,
                       TransformerLevel::Level1);
+    TransformerTester(build_test_case,
+                      check_graph,
+                      TransformerLevel::Default,
+                      TransformerLevel::Level1,
+                      21);
   };
 
   test_case({1, 13, 13, 23}, 4, {0, 3, 1, 2}, false, false, false /*use_contrib_qdq*/);
@@ -3723,7 +3728,12 @@ TEST(QDQTransformerTests, QDQPropagation_DQForward) {
                       TransformerLevel::Level1,
                       18, 0.0, 0.0, nullptr, {},  // defaults that we're not overriding
                       {"TransposeOptimizer"});    // disable TransposeOptimizer for simplicity
-    // TODO: fix opset 19
+    TransformerTester(build_test_case,
+                      check_graph,
+                      TransformerLevel::Default,
+                      TransformerLevel::Level1,
+                      21, 0.0, 0.0, nullptr, {},  // defaults that we're not overriding
+                      {"TransposeOptimizer"});    // disable TransposeOptimizer for simplicity
   };
 
   test_case({1, 13, 13, 23}, 4, {0, 3, 1, 2}, false, false, false /*use_contrib_qdq*/);
diff --git a/onnxruntime/test/optimizer/transpose_optimizer_test.cc b/onnxruntime/test/optimizer/transpose_optimizer_test.cc
index 3cc7eb8f3675f..94383c5de99ba 100644
--- a/onnxruntime/test/optimizer/transpose_optimizer_test.cc
+++ b/onnxruntime/test/optimizer/transpose_optimizer_test.cc
@@ -123,7 +123,7 @@ TEST(TransposeOptimizerTests, TestSplit) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSplitDefaultAxis) {
@@ -156,7 +156,7 @@ TEST(TransposeOptimizerTests, TestSplitDefaultAxis) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSplitNegativeAxis) {
@@ -190,7 +190,7 @@ TEST(TransposeOptimizerTests, TestSplitNegativeAxis) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestConcat) {
@@ -221,7 +221,7 @@ TEST(TransposeOptimizerTests, TestConcat) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestPad) {
@@ -257,7 +257,7 @@ TEST(TransposeOptimizerTests, TestPad) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {10, 18});
+                    /*opset_version*/ {10, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestPadOpset15) {
@@ -286,7 +286,7 @@ TEST(TransposeOptimizerTests, TestPadOpset15) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestPadNonconst) {
@@ -318,7 +318,7 @@ TEST(TransposeOptimizerTests, TestPadNonconst) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {11, 18});
+                    /*opset_version*/ {11, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestResize) {
@@ -382,7 +382,7 @@ TEST(TransposeOptimizerTests, TestResizeOpset11) {
                     // need the level 2 TransposeOptimizer as pushing a Transpose through a Resize requires it to be
                     // assigned to the CPU EP first
                     TransformerLevel::Level2,
-                    /*opset_version*/ {11, 18});
+                    /*opset_version*/ {11, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestResizeOpset15) {
@@ -412,7 +412,7 @@ TEST(TransposeOptimizerTests, TestResizeOpset15) {
                     // need the level 2 TransposeOptimizer as pushing a Transpose through a Resize requires it to be
                     // assigned to the CPU EP first
                     TransformerLevel::Level2,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestResizeSizeRoi) {
@@ -444,7 +444,7 @@ TEST(TransposeOptimizerTests, TestResizeSizeRoi) {
                     // need the level 2 TransposeOptimizer as pushing a Transpose through a Resize requires it to be
                     // assigned to the CPU EP first
                     TransformerLevel::Level2,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestResizeRoiScalesZeroRank0) {
@@ -511,7 +511,7 @@ TEST(TransposeOptimizerTests, TestResizeNonconst) {
                     // need the level 2 TransposeOptimizer as pushing a Transpose through a Resize requires it to be
                     // assigned to the CPU EP first
                     TransformerLevel::Level2,
-                    /*opset_version*/ {11, 18});
+                    /*opset_version*/ {11, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestResizeNonconstOpset13) {
@@ -542,7 +542,7 @@ TEST(TransposeOptimizerTests, TestResizeNonconstOpset13) {
                     // need the level 2 TransposeOptimizer as pushing a Transpose through a Resize requires it to be
                     // assigned to the CPU EP first
                     TransformerLevel::Level2,
-                    /*opset_version*/ {13, 18});
+                    /*opset_version*/ {13, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestAdd) {
@@ -569,7 +569,7 @@ TEST(TransposeOptimizerTests, TestAdd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShape) {
@@ -615,7 +615,7 @@ TEST(TransposeOptimizerTests, TestShapeOpset15) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShapeSliceNoStart) {
@@ -639,7 +639,7 @@ TEST(TransposeOptimizerTests, TestShapeSliceNoStart) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShapeSliceNegativeEnd) {
@@ -663,7 +663,7 @@ TEST(TransposeOptimizerTests, TestShapeSliceNegativeEnd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShapeSliceNegativeStartNoEnd) {
@@ -687,7 +687,7 @@ TEST(TransposeOptimizerTests, TestShapeSliceNegativeStartNoEnd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShapeSliceStartAndEnd) {
@@ -712,7 +712,7 @@ TEST(TransposeOptimizerTests, TestShapeSliceStartAndEnd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestShapeSliceEmptyResult) {
@@ -737,7 +737,7 @@ TEST(TransposeOptimizerTests, TestShapeSliceEmptyResult) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceSumKeepdimsTrue) {
@@ -771,7 +771,7 @@ TEST(TransposeOptimizerTests, TestReduceSumKeepdimsTrue) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {7, 18},
+                    /*opset_version*/ {7, 18, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -797,7 +797,7 @@ TEST(TransposeOptimizerTests, TestReduceSumEmptyAxesKeepdimsTrue) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {7, 18},
+                    /*opset_version*/ {7, 18, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -833,7 +833,7 @@ TEST(TransposeOptimizerTests, TestReduceSumKeepdimsFalse) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {7, 18},
+                    /*opset_version*/ {7, 18, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -859,7 +859,7 @@ TEST(TransposeOptimizerTests, TestReduceSumEmptyAxesKeepdimsFalse) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {7, 18},
+                    /*opset_version*/ {7, 18, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -1235,7 +1235,7 @@ TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsTrue) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsTrueDefaultAxes) {
@@ -1259,7 +1259,7 @@ TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsTrueDefaultAxes) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsFalse) {
@@ -1293,7 +1293,7 @@ TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsFalse) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsFalseDefaultAxes) {
@@ -1317,7 +1317,7 @@ TEST(TransposeOptimizerTests, TestReduceMaxKeepdimsFalseDefaultAxes) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceMax) {
@@ -1349,7 +1349,7 @@ TEST(TransposeOptimizerTests, TestReduceMax) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceMaxDefaultAxes) {
@@ -1372,7 +1372,7 @@ TEST(TransposeOptimizerTests, TestReduceMaxDefaultAxes) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceLogSum) {
@@ -1406,7 +1406,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceLogSum) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceLogSumExp) {
@@ -1440,7 +1440,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceLogSumExp) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceMax) {
@@ -1474,7 +1474,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceMax) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceMean) {
@@ -1508,7 +1508,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceMean) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceMin) {
@@ -1542,7 +1542,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceMin) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceProd) {
@@ -1576,7 +1576,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceProd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceSumSquare) {
@@ -1610,7 +1610,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceSumSquare) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceL1) {
@@ -1644,7 +1644,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceL1) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReduceOpsReduceL2) {
@@ -1678,7 +1678,7 @@ TEST(TransposeOptimizerTests, TestReduceOpsReduceL2) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSqueezeOpset7) {
@@ -1781,7 +1781,7 @@ TEST(TransposeOptimizerTests, TestSqueezeEmptyNoOpt) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {7, 18});
+                    /*opset_version*/ {7, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSqueezeEmptyNoOptOpset15) {
@@ -1826,7 +1826,7 @@ TEST(TransposeOptimizerTests, TestSqueezeNonconstNoOpt) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestUnsqueezeOpset7) {
@@ -1933,7 +1933,7 @@ TEST(TransposeOptimizerTests, TestUnsqueezeNonconstNoOpt) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 14);
+                    /*opset_version*/ {14, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSlice) {
@@ -2019,7 +2019,7 @@ TEST(TransposeOptimizerTests, TestSliceOpset15) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceNoAxesOpset15) {
@@ -2047,7 +2047,7 @@ TEST(TransposeOptimizerTests, TestSliceNoAxesOpset15) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceNegativeAxesInt32) {
@@ -2076,7 +2076,7 @@ TEST(TransposeOptimizerTests, TestSliceNegativeAxesInt32) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceStepsInt32) {
@@ -2106,7 +2106,7 @@ TEST(TransposeOptimizerTests, TestSliceStepsInt32) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceNegativeAxes) {
@@ -2135,7 +2135,7 @@ TEST(TransposeOptimizerTests, TestSliceNegativeAxes) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceSteps) {
@@ -2165,7 +2165,7 @@ TEST(TransposeOptimizerTests, TestSliceSteps) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceNonconstNoOpt) {
@@ -2193,7 +2193,7 @@ TEST(TransposeOptimizerTests, TestSliceNonconstNoOpt) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceNonconstInt32NoOpt) {
@@ -2221,7 +2221,7 @@ TEST(TransposeOptimizerTests, TestSliceNonconstInt32NoOpt) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStarts) {
@@ -2249,7 +2249,7 @@ TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStarts) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsUnknownLengthNoOpt) {
@@ -2276,7 +2276,7 @@ TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsUnknownLengthNoO
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsInt32) {
@@ -2304,7 +2304,7 @@ TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsInt32) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsUnknownLengthInt32NoOpt) {
@@ -2331,7 +2331,7 @@ TEST(TransposeOptimizerTests, TestSliceDefaultAxesNonconstStartsUnknownLengthInt
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestTile) {
@@ -2358,7 +2358,7 @@ TEST(TransposeOptimizerTests, TestTile) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestTileNonconstReps) {
@@ -2385,7 +2385,7 @@ TEST(TransposeOptimizerTests, TestTileNonconstReps) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMinNoAxisKeepdimsTrue) {
@@ -2412,7 +2412,7 @@ TEST(TransposeOptimizerTests, TestArgMinNoAxisKeepdimsTrue) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMinNoAxisKeepdimsFalse) {
@@ -2439,7 +2439,7 @@ TEST(TransposeOptimizerTests, TestArgMinNoAxisKeepdimsFalse) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMinNoAxis) {
@@ -2465,7 +2465,7 @@ TEST(TransposeOptimizerTests, TestArgMinNoAxis) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMinKeepdimsTrue) {
@@ -2493,7 +2493,7 @@ TEST(TransposeOptimizerTests, TestArgMinKeepdimsTrue) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMinKeepdimsFalse) {
@@ -2521,7 +2521,7 @@ TEST(TransposeOptimizerTests, TestArgMinKeepdimsFalse) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMin) {
@@ -2548,7 +2548,7 @@ TEST(TransposeOptimizerTests, TestArgMin) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestArgMax) {
@@ -2576,7 +2576,7 @@ TEST(TransposeOptimizerTests, TestArgMax) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSoftmax) {
@@ -2603,7 +2603,7 @@ TEST(TransposeOptimizerTests, TestSoftmax) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2631,7 +2631,7 @@ TEST(TransposeOptimizerTests, TestSoftmaxNoAxis) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2660,7 +2660,7 @@ TEST(TransposeOptimizerTests, TestSoftmax_2) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2688,7 +2688,7 @@ TEST(TransposeOptimizerTests, TestSoftmaxNoOptimization) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2716,7 +2716,7 @@ TEST(TransposeOptimizerTests, TestSoftmaxNoOptimization_2) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2744,7 +2744,7 @@ TEST(TransposeOptimizerTests, TestSoftmaxNoOptimization_3) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 12,
+                    /*opset_version*/ {12, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2829,7 +2829,7 @@ TEST(TransposeOptimizerTests, TestHardmaxAndLogSoftmaxNoAxis) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 15,
+                    /*opset_version*/ {15, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2857,7 +2857,7 @@ TEST(TransposeOptimizerTests, TestHardmaxAndLogSoftmaxNoAxis_2) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 15,
+                    /*opset_version*/ {15, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 }
@@ -2889,7 +2889,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsAdd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsMul) {
@@ -2919,7 +2919,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsMul) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsSub) {
@@ -2949,7 +2949,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsSub) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsDiv) {
@@ -2982,7 +2982,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsDiv) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18},
+                    /*opset_version*/ {15, 18, 23},
                     /*per_sample_tolerance*/ 1e-07,
                     /*relative_per_sample_tolerance*/ 1e-06);
 #else
@@ -2990,7 +2990,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsDiv) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 #endif  // defined(_M_ARM64) && _MSC_VER >= 1930
 }
 
@@ -3021,7 +3021,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsPRelu) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsGreater) {
@@ -3051,7 +3051,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsGreater) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsLess) {
@@ -3081,7 +3081,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsLess) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsPow) {
@@ -3111,7 +3111,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsPow) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsMax) {
@@ -3141,7 +3141,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsMax) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsMin) {
@@ -3171,7 +3171,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsMin) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsMean) {
@@ -3201,7 +3201,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsMean) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsSum) {
@@ -3231,7 +3231,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsSum) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsGreaterOrEqual) {
@@ -3261,7 +3261,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsGreaterOrEqual) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsLessOrEqual) {
@@ -3291,7 +3291,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsLessOrEqual) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsEqual) {
@@ -3321,7 +3321,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsEqual) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsAnd) {
@@ -3351,7 +3351,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsAnd) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsOr) {
@@ -3381,7 +3381,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsOr) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsXor) {
@@ -3411,7 +3411,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsXor) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsMod) {
@@ -3442,7 +3442,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsMod) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastOpsBitShift) {
@@ -3473,7 +3473,7 @@ TEST(TransposeOptimizerTests, TestBroadcastOpsBitShift) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestWhere) {
@@ -3504,7 +3504,7 @@ TEST(TransposeOptimizerTests, TestWhere) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 // Utility function that runs TransformerTester for the graph Transpose -> QuantizeLinear -> Transpose.
@@ -3797,7 +3797,7 @@ TEST(TransposeOptimizerTests, TestCast) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestQLinearSoftmax) {
@@ -3827,17 +3827,19 @@ TEST(TransposeOptimizerTests, TestQLinearSoftmax) {
     EXPECT_EQ(transpose_cost, 0);
   };
 
-  TransformerTester(build_test_case_1,
-                    check_optimized_graph_1,
-                    TransformerLevel::Level2,
-                    TransformerLevel::Level3,
-                    /*opset_version*/ 13,
-                    /*per_sample_tolerance*/ 0.0,
-                    /*relative_per_sample_tolerance*/ 0.0,
-                    /*transformer*/ nullptr,
-                    /*add_session_options*/ {},
-                    /*disabled_optimizers*/ {},
-                    /*ep*/ DefaultCpuExecutionProvider());
+  for (int opset_version : {13, 23}) {
+    TransformerTester(build_test_case_1,
+                      check_optimized_graph_1,
+                      TransformerLevel::Level2,
+                      TransformerLevel::Level3,
+                      /*opset_version*/ opset_version,
+                      /*per_sample_tolerance*/ 0.0,
+                      /*relative_per_sample_tolerance*/ 0.0,
+                      /*transformer*/ nullptr,
+                      /*add_session_options*/ {},
+                      /*disabled_optimizers*/ {},
+                      /*ep*/ DefaultCpuExecutionProvider());
+  }
 }
 
 TEST(TransposeOptimizerTests, TestBroadcastReusedInputs) {
@@ -3868,7 +3870,7 @@ TEST(TransposeOptimizerTests, TestBroadcastReusedInputs) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestTransposeGraphOutput) {
@@ -3896,7 +3898,7 @@ TEST(TransposeOptimizerTests, TestTransposeGraphOutput) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestSimpleReshapeAsTranspose) {
@@ -3929,7 +3931,7 @@ TEST(TransposeOptimizerTests, TestSimpleReshapeAsTranspose) {
                     check_optimized_graph,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestReshapeAsTransposeGraphOutput) {
@@ -3960,7 +3962,7 @@ TEST(TransposeOptimizerTests, TestReshapeAsTransposeGraphOutput) {
                     check_optimized_graph,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 enum class TransposeReshapeResult {
@@ -4029,7 +4031,7 @@ static void TestTransposeReshape(const std::vector<int64_t>& input_shape,    //
                     check_optimized_graph,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ 15);
+                    /*opset_version*/ {15, 23});
 }
 
 // Transpose -> Reshape can be merged if the Reshape could also be expressed as a Transpose due to not changing the
@@ -4180,7 +4182,7 @@ TEST(TransposeOptimizerTests, TestCancelingNodesGraphOutputs) {
                     check_optimized_graph,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestNonCancelingReshapeDueToNonConstShape) {
@@ -4216,7 +4218,7 @@ TEST(TransposeOptimizerTests, TestNonCancelingReshapeDueToNonConstShape) {
                     check_optimized_graph,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestPushBroadcastUnsqueezeTranspose) {
@@ -4251,7 +4253,7 @@ TEST(TransposeOptimizerTests, TestPushBroadcastUnsqueezeTranspose) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestOptimizeTowardsTranspose) {
@@ -4281,7 +4283,7 @@ TEST(TransposeOptimizerTests, TestOptimizeTowardsTranspose) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestOnlyOptimizeTowardsTranspose) {
@@ -4308,7 +4310,7 @@ TEST(TransposeOptimizerTests, TestOnlyOptimizeTowardsTranspose) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestDontOptimizeWrongInput) {
@@ -4334,7 +4336,7 @@ TEST(TransposeOptimizerTests, TestDontOptimizeWrongInput) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestOptimizeBothInputs) {
@@ -4362,7 +4364,7 @@ TEST(TransposeOptimizerTests, TestOptimizeBothInputs) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 TEST(TransposeOptimizerTests, TestOmitIdentityTranspose) {
@@ -4394,7 +4396,7 @@ TEST(TransposeOptimizerTests, TestOmitIdentityTranspose) {
                     check_optimized_graph_1,
                     TransformerLevel::Default,
                     TransformerLevel::Level1,
-                    /*opset_version*/ {15, 18});
+                    /*opset_version*/ {15, 18, 23});
 }
 
 // regression test for a model where the transpose optimizations were not completed in a single pass in level 1.