diff --git a/tensorflow/lite/delegates/gpu/common/gpu_info.cc b/tensorflow/lite/delegates/gpu/common/gpu_info.cc
index 63cdec01e57e2c..3ac82064cc8175 100644
--- a/tensorflow/lite/delegates/gpu/common/gpu_info.cc
+++ b/tensorflow/lite/delegates/gpu/common/gpu_info.cc
@@ -557,6 +557,14 @@ bool OpenGlInfo::SupportsExplicitFp16() const {
   return supports_f16_alu && supports_f16_storage;
 }
 
+bool OpenGlInfo::IsApiOpenGl31OrAbove() const {
+  return (major_version == 3 && minor_version >= 1) || major_version > 3;
+}
+
+bool OpenGlInfo::IsApiOpenGl32OrAbove() const {
+  return (major_version == 3 && minor_version >= 2) || major_version > 3;
+}
+
 bool VulkanInfo::SupportsExplicitFp16() const {
   bool supports_f16_alu = false;
   bool supports_f16_storage = false;
@@ -691,6 +699,28 @@ bool GpuInfo::SupportsPointersInKernels() const {
   return IsApiOpenCl() || IsApiMetal();
 }
 
+bool GpuInfo::SupportsZeroClampForImageBuffer() const {
+  if (IsApiMetal() || IsApiOpenCl()) {
+    return true;
+  } else {
+    return false;
+  }
+}
+
+bool GpuInfo::SupportsZeroClampForImages() const {
+  if (IsApiMetal()) {
+    return true;
+  } else if (IsApiOpenCl()) {
+    return true;
+  } else if (IsApiVulkan()) {
+    return true;
+  } else if (IsApiOpenGl()) {
+    return opengl_info.IsApiOpenGl32OrAbove();
+  } else {
+    return false;
+  }
+}
+
 bool GpuInfo::IsWaveSizeEqualTo32() const {
   return supported_subgroup_sizes.size() == 1 &&
          supported_subgroup_sizes[0] == 32;
@@ -949,8 +979,7 @@ bool GpuInfo::IsApiOpenGl31OrAbove() const {
   if (!IsApiOpenGl()) {
     return false;
   }
-  return (opengl_info.major_version == 3 && opengl_info.minor_version >= 1) ||
-         opengl_info.major_version > 3;
+  return opengl_info.IsApiOpenGl31OrAbove();
 }
 
 bool GpuInfo::IsApiVulkan() const { return gpu_api == GpuApi::kVulkan; }
diff --git a/tensorflow/lite/delegates/gpu/common/gpu_info.h b/tensorflow/lite/delegates/gpu/common/gpu_info.h
index 69dc42d21a95c2..8a5ddc2f2b662b 100644
--- a/tensorflow/lite/delegates/gpu/common/gpu_info.h
+++ b/tensorflow/lite/delegates/gpu/common/gpu_info.h
@@ -278,6 +278,9 @@ struct OpenGlInfo {
   int max_compute_work_group_size_z;
 
   bool SupportsExplicitFp16() const;
+
+  bool IsApiOpenGl31OrAbove() const;
+  bool IsApiOpenGl32OrAbove() const;
 };
 
 struct VulkanInfo {
@@ -437,6 +440,9 @@ struct GpuInfo {
   bool SupportsFloatImage2D(DataType data_type, int channels) const;
   bool SupportsExtension(const std::string& extension) const;
 
+  bool SupportsZeroClampForImageBuffer() const;
+  bool SupportsZeroClampForImages() const;
+
   int GetComputeUnitsCount() const;
 
   int GetMaxImageArguments() const;
diff --git a/tensorflow/lite/delegates/gpu/common/gpu_model.cc b/tensorflow/lite/delegates/gpu/common/gpu_model.cc
index cce79214ab3aba..b2fc8a44d2c05f 100644
--- a/tensorflow/lite/delegates/gpu/common/gpu_model.cc
+++ b/tensorflow/lite/delegates/gpu/common/gpu_model.cc
@@ -197,6 +197,12 @@ absl::Status ReserveGraphTensors(const CreateGpuModelInfo& create_info,
                                  const GpuInfo& gpu_info,
                                  const GraphFloat32& graph,
                                  TensorReserver* tensor_reserver) {
+  ZeroClampSupport zero_clamp_support;
+  zero_clamp_support.image_buffer = gpu_info.SupportsZeroClampForImageBuffer();
+  zero_clamp_support.image2d = gpu_info.SupportsZeroClampForImages();
+  zero_clamp_support.image2d_array = gpu_info.SupportsZeroClampForImages();
+  zero_clamp_support.image3d = gpu_info.SupportsZeroClampForImages();
+
   ValueId max_id = 0;
   auto tensors = graph.values();
   for (auto& t : tensors) {
@@ -261,6 +267,7 @@ absl::Status ReserveGraphTensors(const CreateGpuModelInfo& create_info,
       }
     }
     tensor_desc.SetBHWCShape(shape);
+    tensor_desc.SetZeroClampSupport(zero_clamp_support);
     tensor_reserver->Add(t->id, tensor_desc);
     max_id = std::max(max_id, t->id);
   }