diff --git a/mlir/include/mlir/Conversion/GPUCommon/GPUCommonPass.h b/mlir/include/mlir/Conversion/GPUCommon/GPUCommonPass.h
--- a/mlir/include/mlir/Conversion/GPUCommon/GPUCommonPass.h
+++ b/mlir/include/mlir/Conversion/GPUCommon/GPUCommonPass.h
@@ -26,6 +26,9 @@
 
 namespace gpu {
 class GPUModuleOp;
+
+/// Returns the default annotation name for GPU binary blobs.
+std::string getDefaultGpuBinaryAnnotation();
 } // namespace gpu
 
 namespace LLVM {
@@ -44,13 +47,17 @@
 /// This pass does not generate code to call GPU runtime APIs directly but
 /// instead uses a small wrapper library that exports a stable and conveniently
 /// typed ABI on top of GPU runtimes such as CUDA or ROCm (HIP).
+///
+/// A non-empty gpuBinaryAnnotation overrides the pass' command line option.
 std::unique_ptr<OperationPass<ModuleOp>>
-createGpuToLLVMConversionPass(StringRef gpuBinaryAnnotation = "");
+createGpuToLLVMConversionPass(StringRef gpuBinaryAnnotation = {});
 
 /// Collect a set of patterns to convert from the GPU dialect to LLVM.
+///
+/// A non-empty gpuBinaryAnnotation overrides the pass' command line option.
 void populateGpuToLLVMConversionPatterns(LLVMTypeConverter &converter,
                                          OwningRewritePatternList &patterns,
-                                         StringRef gpuBinaryAnnotation);
+                                         StringRef gpuBinaryAnnotation = {});
 
 /// Creates a pass to convert kernel functions into GPU target object blobs.
 ///
@@ -74,11 +81,13 @@
 ///
 /// After the transformation, the body of the kernel function is removed (i.e.,
 /// it is turned into a declaration).
+///
+/// A non-empty gpuBinaryAnnotation overrides the pass' command line option.
 std::unique_ptr<OperationPass<gpu::GPUModuleOp>>
 createConvertGPUKernelToBlobPass(LoweringCallback loweringCallback,
                                  BlobGenerator blobGenerator, StringRef triple,
                                  StringRef targetChip, StringRef features,
-                                 StringRef gpuBinaryAnnotation);
+                                 StringRef gpuBinaryAnnotation = {});
 
 } // namespace mlir
 
diff --git a/mlir/include/mlir/Conversion/Passes.td b/mlir/include/mlir/Conversion/Passes.td
--- a/mlir/include/mlir/Conversion/Passes.td
+++ b/mlir/include/mlir/Conversion/Passes.td
@@ -106,10 +106,6 @@
   let summary = "Convert GPU dialect to LLVM dialect with GPU runtime calls";
   let constructor = "mlir::createGpuToLLVMConversionPass()";
   let dependentDialects = ["LLVM::LLVMDialect"];
-  let options = [
-    Option<"gpuBinaryAnnotation", "gpu-binary-annotation", "std::string",
-           "", "Annotation attribute string for GPU binary">,
-  ];
 }
 
 def LowerHostCodeToLLVM : Pass<"lower-host-to-llvm", "ModuleOp"> {
diff --git a/mlir/lib/Conversion/GPUCommon/ConvertKernelFuncToBlob.cpp b/mlir/lib/Conversion/GPUCommon/ConvertKernelFuncToBlob.cpp
--- a/mlir/lib/Conversion/GPUCommon/ConvertKernelFuncToBlob.cpp
+++ b/mlir/lib/Conversion/GPUCommon/ConvertKernelFuncToBlob.cpp
@@ -52,8 +52,15 @@
                       StringRef targetChip, StringRef features,
                       StringRef gpuBinaryAnnotation)
       : loweringCallback(loweringCallback), blobGenerator(blobGenerator),
-        triple(triple), targetChip(targetChip), features(features),
-        blobAnnotation(gpuBinaryAnnotation) {}
+        triple(triple), targetChip(targetChip), features(features) {
+    if (!gpuBinaryAnnotation.empty())
+      this->gpuBinaryAnnotation = gpuBinaryAnnotation.str();
+  }
+
+  GpuKernelToBlobPass(const GpuKernelToBlobPass &other)
+      : loweringCallback(other.loweringCallback),
+        blobGenerator(other.blobGenerator), triple(other.triple),
+        targetChip(other.targetChip), features(other.features) {}
 
   void runOnOperation() override {
     gpu::GPUModuleOp module = getOperation();
@@ -70,7 +77,7 @@
     // attribute to the module.
     if (auto blobAttr = translateGPUModuleToBinaryAnnotation(
             *llvmModule, module.getLoc(), module.getName()))
-      module->setAttr(blobAnnotation, blobAttr);
+      module->setAttr(gpuBinaryAnnotation, blobAttr);
     else
       signalPassFailure();
   }
@@ -92,14 +99,21 @@
 
   LoweringCallback loweringCallback;
   BlobGenerator blobGenerator;
+
   llvm::Triple triple;
   std::string targetChip;
   std::string features;
-  std::string blobAnnotation;
+
+  Option<std::string> gpuBinaryAnnotation{
+      *this, "gpu-binary-annotation",
+      llvm::cl::desc("Annotation attribute string for GPU binary"),
+      llvm::cl::init(gpu::getDefaultGpuBinaryAnnotation())};
 };
 
 } // anonymous namespace
 
+std::string gpu::getDefaultGpuBinaryAnnotation() { return "gpu.binary"; }
+
 std::string
 GpuKernelToBlobPass::translateModuleToISA(llvm::Module &module,
                                           llvm::TargetMachine &targetMachine) {
diff --git a/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp b/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
--- a/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
+++ b/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
@@ -48,8 +48,17 @@
       this->gpuBinaryAnnotation = gpuBinaryAnnotation.str();
   }
 
+  GpuToLLVMConversionPass(const GpuToLLVMConversionPass &other)
+      : GpuToLLVMConversionPassBase(other) {}
+
   // Run the dialect converter on the module.
   void runOnOperation() override;
+
+private:
+  Option<std::string> gpuBinaryAnnotation{
+      *this, "gpu-binary-annotation",
+      llvm::cl::desc("Annotation attribute string for GPU binary"),
+      llvm::cl::init(gpu::getDefaultGpuBinaryAnnotation())};
 };
 
 struct FunctionCallBuilder {
diff --git a/mlir/tools/mlir-cuda-runner/mlir-cuda-runner.cpp b/mlir/tools/mlir-cuda-runner/mlir-cuda-runner.cpp
--- a/mlir/tools/mlir-cuda-runner/mlir-cuda-runner.cpp
+++ b/mlir/tools/mlir-cuda-runner/mlir-cuda-runner.cpp
@@ -114,7 +114,8 @@
     : public mlir::PassPipelineOptions<GpuToCubinPipelineOptions> {
   Option<std::string> gpuBinaryAnnotation{
       *this, "gpu-binary-annotation",
-      llvm::cl::desc("Annotation attribute string for GPU binary")};
+      llvm::cl::desc("Annotation attribute string for GPU binary"),
+      llvm::cl::init(gpu::getDefaultGpuBinaryAnnotation())};
 };
 
 // Register cuda-runner specific passes.