diff --git a/mediapipe/calculators/tensor/inference_calculator_gl_advanced.cc b/mediapipe/calculators/tensor/inference_calculator_gl_advanced.cc
index 8aee46185..e265eaee7 100644
--- a/mediapipe/calculators/tensor/inference_calculator_gl_advanced.cc
+++ b/mediapipe/calculators/tensor/inference_calculator_gl_advanced.cc
@@ -69,6 +69,7 @@ class InferenceCalculatorGlAdvancedImpl
             gpu_delegate_options);
     absl::Status ReadGpuCaches(tflite::gpu::TFLiteGPURunner* gpu_runner) const;
     absl::Status SaveGpuCaches(tflite::gpu::TFLiteGPURunner* gpu_runner) const;
+    bool UseSerializedModel() const { return use_serialized_model_; }
 
    private:
     bool use_kernel_caching_ = false;
@@ -150,8 +151,6 @@ InferenceCalculatorGlAdvancedImpl::GpuInferenceRunner::Process(
 }
 
 absl::Status InferenceCalculatorGlAdvancedImpl::GpuInferenceRunner::Close() {
-  MP_RETURN_IF_ERROR(
-      on_disk_cache_helper_.SaveGpuCaches(tflite_gpu_runner_.get()));
   return gpu_helper_.RunInGlContext([this]() -> absl::Status {
     tflite_gpu_runner_.reset();
     return absl::OkStatus();
@@ -226,9 +225,14 @@ InferenceCalculatorGlAdvancedImpl::GpuInferenceRunner::InitTFLiteGPURunner(
                          tflite_gpu_runner_->GetOutputShapes()[i].c};
   }
 
+  if (on_disk_cache_helper_.UseSerializedModel()) {
+    tflite_gpu_runner_->ForceOpenCLInitFromSerializedModel();
+  }
+
   MP_RETURN_IF_ERROR(
       on_disk_cache_helper_.ReadGpuCaches(tflite_gpu_runner_.get()));
-  return tflite_gpu_runner_->Build();
+  MP_RETURN_IF_ERROR(tflite_gpu_runner_->Build());
+  return on_disk_cache_helper_.SaveGpuCaches(tflite_gpu_runner_.get());
 }
 
 #if defined(MEDIAPIPE_ANDROID) || defined(MEDIAPIPE_CHROMIUMOS)
diff --git a/mediapipe/util/tflite/tflite_gpu_runner.cc b/mediapipe/util/tflite/tflite_gpu_runner.cc
index 4e40975cb..c1b272b67 100644
--- a/mediapipe/util/tflite/tflite_gpu_runner.cc
+++ b/mediapipe/util/tflite/tflite_gpu_runner.cc
@@ -234,6 +234,11 @@ absl::Status TFLiteGPURunner::InitializeOpenCL(
   MP_RETURN_IF_ERROR(
       cl::NewInferenceEnvironment(env_options, &cl_environment_, &properties));
 
+  if (serialized_model_.empty() &&
+      opencl_init_from_serialized_model_is_forced_) {
+    ASSIGN_OR_RETURN(serialized_model_, GetSerializedModel());
+  }
+
   // Try to initialize from serialized model first.
   if (!serialized_model_.empty()) {
     absl::Status init_status = InitializeOpenCLFromSerializedModel(builder);
@@ -270,7 +275,6 @@ absl::Status TFLiteGPURunner::InitializeOpenCLFromSerializedModel(
 }
 
 absl::StatusOr<std::vector<uint8_t>> TFLiteGPURunner::GetSerializedModel() {
-  RET_CHECK(runner_) << "Runner is in invalid state.";
   if (serialized_model_used_) {
     return serialized_model_;
   }
diff --git a/mediapipe/util/tflite/tflite_gpu_runner.h b/mediapipe/util/tflite/tflite_gpu_runner.h
index 5eeaa230f..c64981ef8 100644
--- a/mediapipe/util/tflite/tflite_gpu_runner.h
+++ b/mediapipe/util/tflite/tflite_gpu_runner.h
@@ -62,6 +62,9 @@ class TFLiteGPURunner {
 
   void ForceOpenGL() { opengl_is_forced_ = true; }
   void ForceOpenCL() { opencl_is_forced_ = true; }
+  void ForceOpenCLInitFromSerializedModel() {
+    opencl_init_from_serialized_model_is_forced_ = true;
+  }
 
   absl::Status BindSSBOToInputTensor(GLuint ssbo_id, int input_id);
   absl::Status BindSSBOToOutputTensor(GLuint ssbo_id, int output_id);
@@ -141,6 +144,7 @@ class TFLiteGPURunner {
 
   bool opencl_is_forced_ = false;
   bool opengl_is_forced_ = false;
+  bool opencl_init_from_serialized_model_is_forced_ = false;
 };
 
 }  // namespace gpu