diff --git a/csrc/id_model/indexing.cpp b/csrc/id_model/indexing.cpp
index 4c42da75010..9d71a0c3e3e 100644
--- a/csrc/id_model/indexing.cpp
+++ b/csrc/id_model/indexing.cpp
@@ -34,6 +34,8 @@
 namespace nvfuser {
 
 TensorIndexer::TensorIndexer(IdModel& id_model) : id_model_(id_model) {
+  NVF_ERROR(isSupported(id_model.fusion()));
+
   buildLoopIndexMap();
 
   if (isDebugDumpEnabled(DebugDumpOption::IndexingVerbose)) {
diff --git a/csrc/scheduler/expr_eval_sched.cpp b/csrc/scheduler/expr_eval_sched.cpp
index f2b6ebce41b..f540e545bb5 100644
--- a/csrc/scheduler/expr_eval_sched.cpp
+++ b/csrc/scheduler/expr_eval_sched.cpp
@@ -64,6 +64,7 @@ bool ExprEvalScheduler::canScheduleCompileTime(Fusion* fusion) {
   // TODO: remove IndexPutAccumulateOp
   if (exprs.front()
           ->isOneOf<
+              GatherOp,
               ScatterOp,
               SdpaFwdOp,
               SdpaBwdOp,
diff --git a/csrc/scheduler/registry.cpp b/csrc/scheduler/registry.cpp
index 7e47cc253b7..a04714e22b0 100644
--- a/csrc/scheduler/registry.cpp
+++ b/csrc/scheduler/registry.cpp
@@ -64,6 +64,16 @@ bool checkCanSchedule(Fusion* fusion, SchedulerType scheduler_type) {
     return false;
   }
 
+  // Support of non-exact gather was dropped when the legacy indexer was
+  // deprecated
+  if (std::ranges::any_of(
+          ir_utils::getOpsOfType<GatherOp>(fusion),
+          [](GatherOp* gather) { return !gather->exactSizes(); })) {
+    scheduler_debug_utils::canScheduleRejectReason(
+        scheduler_type, "Non-exact gather ops");
+    return false;
+  }
+
   // Fusions with `MatmulOp, LinearOp, MmaOp` can only be accepted by Matmul
   // scheduler.
   if (scheduler_type != SchedulerType::Matmul &&
diff --git a/tests/cpp/test_gather.cpp b/tests/cpp/test_gather.cpp
index 0924b33a70b..92e78f7c4fa 100644
--- a/tests/cpp/test_gather.cpp
+++ b/tests/cpp/test_gather.cpp
@@ -583,7 +583,7 @@ TEST_F(GatherTest, TakeAlongAxisIntermediateTensorReduction1) {
 
   validateSegmentation(
       executor_cache.getMostRecentKernelRuntime(),
-      {SchedulerType::Reduction, SchedulerType::PointWise});
+      {SchedulerType::Reduction, SchedulerType::ExprEval});
 
   testValidate(&fusion, outputs, {t0, t1}, __LINE__, __FILE__);
 }
@@ -1128,7 +1128,8 @@ TEST_F(GatherTest, TakeAlongAxisCrossEntropyLoss) {
 }
 
 // Test grouped reduction on IterType::GatherScatter
-TEST_F(GatherTest, GatherIterGoupedReduction) {
+// Codegen support of non-exact gather dropped
+TEST_F(GatherTest, DISABLED_GatherIterGoupedReduction) {
   const int max_dim_size = 128;
   auto options = at::TensorOptions().dtype(at::kFloat).device(at::kCUDA, 0);
   auto options_i = at::TensorOptions().dtype(at::kLong).device(at::kCUDA, 0);
@@ -1212,7 +1213,8 @@ TEST_F(GatherTest, GatherIterGoupedReduction) {
       lparams);
 }
 
-TEST_F(GatherTest, SameTvUsedAsLookupAndIndex) {
+// Codegen support of non-exact gather dropped
+TEST_F(GatherTest, DISABLED_SameTvUsedAsLookupAndIndex) {
   auto fusion_ptr = std::make_unique<Fusion>();
   Fusion& fusion = *fusion_ptr.get();
   FusionGuard fg(&fusion);