diff --git a/test/main.cpp b/test/main.cpp
index 374eaf84b3..0d49dee908 100644
--- a/test/main.cpp
+++ b/test/main.cpp
@@ -2,18 +2,34 @@
 // Licensed under the MIT License.
 
 #include <iostream>
+#include <string>
 
 #include <gtest/gtest.h>
 
 #include "ort_genai.h"
 
+// Global variable to store custom model base path
+std::string g_custom_model_path;
+
 int main(int argc, char** argv) {
   std::cout << "Generators Utility Library" << std::endl;
+
   std::cout << "Initializing OnnxRuntime... ";
   std::cout.flush();
   try {
     std::cout << "done" << std::endl;
     ::testing::InitGoogleTest(&argc, argv);
+
+    // Parse custom model path argument after InitGoogleTest
+    for (int i = 1; i < argc; ++i) {
+      const std::string arg = argv[i];
+      if (arg == "--model_path" && i + 1 < argc) {
+        g_custom_model_path = argv[++i];
+        std::cout << "Using custom model path: " << g_custom_model_path << std::endl;
+        break;
+      }
+    }
+
     int result = RUN_ALL_TESTS();
     std::cout << "Shutting down OnnxRuntime... ";
     OgaShutdown();
diff --git a/test/model_tests.cpp b/test/model_tests.cpp
index 56ce5696e0..70dd8c2ea4 100644
--- a/test/model_tests.cpp
+++ b/test/model_tests.cpp
@@ -5,6 +5,7 @@
 #include <iostream>
 #include <random>
 #include <filesystem>
+#include <string>
 #include <gtest/gtest.h>
 
 #include "span.h"
@@ -12,6 +13,9 @@
 #include <ort_genai.h>
 #include <gtest/gtest.h>
 
+// External global variable from main.cpp for custom model path
+extern std::string g_custom_model_path;
+
 #ifndef MODEL_PATH
 #define MODEL_PATH "../../test/test_models/"
 #endif
@@ -274,9 +278,12 @@ static const std::pair<const char*, const char*> c_phi3_nvtrt_model_paths[] = {
 };
 
 void Test_GreedySearch_Phi3_NvTensorRtRtx(const char* model_path, const char* model_label) {
+  // Use custom path if provided, otherwise use default
+  std::string resolved_path = g_custom_model_path.empty() ? model_path : g_custom_model_path;
+
   // Skip test if NvTensorRT model is not available
-  if (!std::filesystem::exists(model_path)) {
-    GTEST_SKIP() << "NvTensorRT model not available at: " << model_path;
+  if (!std::filesystem::exists(resolved_path)) {
+    GTEST_SKIP() << "NvTensorRT model not available at: " << resolved_path;
   }
   const std::vector<int64_t> input_ids_shape{1, 19};
   const std::vector<int32_t> input_ids{32006, 887, 526, 263, 8444, 29871, 23869, 20255, 29889, 32007, 32010, 6324, 29892, 1128, 526, 366, 29973, 32007, 32001};
@@ -285,7 +292,7 @@ void Test_GreedySearch_Phi3_NvTensorRtRtx(const char* model_path, const char* mo
   const std::vector<int32_t> expected_output{
       32006, 887, 526, 263, 8444, 29871, 23869, 20255, 29889, 32007, 32010, 6324, 29892, 1128, 526, 366, 29973, 32007, 32001,  // Input tokens (19)
       15043, 29991, 306, 29915, 29885, 2599};
-  auto config = OgaConfig::Create(model_path);
+  auto config = OgaConfig::Create(resolved_path.c_str());
   config->ClearProviders();
   config->AppendProvider("NvTensorRtRtx");
   auto model = OgaModel::Create(*config);
@@ -321,9 +328,12 @@ TEST(ModelTests, GreedySearchPhi3NvTensorRtRtx) {
 }
 
 void Test_OutOfPlaceKvCache_Phi3_NvTensorRtRtx(const char* model_path, const char* model_label) {
+  // Use custom path if provided, otherwise use default
+  std::string resolved_path = g_custom_model_path.empty() ? model_path : g_custom_model_path;
+
   // Skip test if NvTensorRT model is not available
-  if (!std::filesystem::exists(model_path)) {
-    GTEST_SKIP() << "NvTensorRT model not available at: " << model_path;
+  if (!std::filesystem::exists(resolved_path)) {
+    GTEST_SKIP() << "NvTensorRT model not available at: " << resolved_path;
   }
 
   const std::vector<int64_t> input_ids_shape{1, 19};
@@ -336,7 +346,7 @@ void Test_OutOfPlaceKvCache_Phi3_NvTensorRtRtx(const char* model_path, const cha
       32006, 887, 526, 263, 8444, 29871, 23869, 20255, 29889, 32007, 32010, 6324, 29892, 1128, 526, 366, 29973, 32007, 32001,  // Input tokens (19)
       15043, 1554, 13, 16271, 29892, 8733};
 
-  auto config = OgaConfig::Create(model_path);
+  auto config = OgaConfig::Create(resolved_path.c_str());
   config->ClearProviders();
   config->AppendProvider("NvTensorRtRtx");
   auto model = OgaModel::Create(*config);
diff --git a/test/sampling_benchmark.cpp b/test/sampling_benchmark.cpp
index 83090d10b5..e1ec1643f3 100644
--- a/test/sampling_benchmark.cpp
+++ b/test/sampling_benchmark.cpp
@@ -25,6 +25,9 @@
 #define MODEL_PATH "../../test/test_models/"
 #endif
 
+// External global variable from main.cpp for custom model path
+extern std::string g_custom_model_path;
+
 // Defined in sampling_tests.cpp
 void CreateRandomLogits(float* logits, int num_large, int vocab_size, int batch_size, std::mt19937& engine);
 
@@ -109,11 +112,11 @@ void PrintSummary(const std::vector<BenchmarkResult>& results) {
 }
 
 BenchmarkResult RunBenchmark(const BenchmarkParams& params) {
-  const char* model_path = MODEL_PATH "hf-internal-testing/tiny-random-gpt2-fp32";
+  std::string model_path = MODEL_PATH "hf-internal-testing/tiny-random-gpt2-fp32";
   if (strcmp(params.device_type, "NvTensorRtRtx") == 0) {
-    model_path = MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt";
+    model_path = g_custom_model_path.empty() ? MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt" : g_custom_model_path;
   }
-  auto config = OgaConfig::Create(model_path);
+  auto config = OgaConfig::Create(model_path.c_str());
   std::string overlay = R"({ "model": { "vocab_size" : )" + std::to_string(params.vocab_size) + R"( } })";
   config->Overlay(overlay.c_str());
   config->ClearProviders();
@@ -184,7 +187,8 @@ TEST(SamplingBenchmarks, PerformanceTests) {
   device_types.push_back("cuda");
 #endif
   // Add NvTensorRtRtx if model is available
-  if (std::filesystem::exists(MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt")) {
+  std::string resolved_nvtrt_path = g_custom_model_path.empty() ? MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt" : g_custom_model_path;
+  if (std::filesystem::exists(resolved_nvtrt_path)) {
     device_types.push_back("NvTensorRtRtx");
   }
 
diff --git a/test/sampling_tests.cpp b/test/sampling_tests.cpp
index 6fced8ed1d..e5de94fab6 100644
--- a/test/sampling_tests.cpp
+++ b/test/sampling_tests.cpp
@@ -19,6 +19,9 @@
 #define MODEL_PATH "../../test/test_models/"
 #endif
 
+// External global variable from main.cpp for custom model path
+extern std::string g_custom_model_path;
+
 TEST(SamplingTests, BatchedSamplingTopPCpu) {
   std::vector<int32_t> input_ids{0, 1, 2, 3};
   std::vector<int32_t> expected_output{1, 2, 3, 4};
@@ -555,8 +558,12 @@ struct NvTensorRtRtxTestSetup {
   static NvTensorRtRtxTestSetup Create(int vocab_size, int batch_size, int max_length = 10) {
     NvTensorRtRtxTestSetup setup;
 
+    // Use custom path if provided, otherwise use default
+    std::string nvtrt_path = MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt";
+    std::string resolved_path = g_custom_model_path.empty() ? nvtrt_path : g_custom_model_path;
+
     // Check if model is available
-    if (!std::filesystem::exists(MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt")) {
+    if (!std::filesystem::exists(resolved_path)) {
       setup.is_available = false;
       return setup;
     }
@@ -564,7 +571,7 @@ struct NvTensorRtRtxTestSetup {
     setup.is_available = true;
 
     // Create config with vocab_size overlay
-    auto config = OgaConfig::Create(MODEL_PATH "hf-internal-testing/phi3-fp16-nvtrt");
+    auto config = OgaConfig::Create(resolved_path.c_str());
     std::string overlay = R"({ "model": { "vocab_size" : )" + std::to_string(vocab_size) + R"( } })";
     config->Overlay(overlay.c_str());
     config->ClearProviders();