Baiju Meswani (baijumeswani) · Sep 14, 2026 · Sep 13, 2026 · Sep 13, 2026 · Sep 13, 2026 · Sep 13, 2026
diff --git a/.github/instructions/cpp-ci-test-policy.instructions.md b/.github/instructions/cpp-ci-test-policy.instructions.md
@@ -34,11 +34,9 @@ unload, not CUDA kernel correctness or trained-model quality.
 
 The packaging pipeline includes the unconditional `cpp_test_engine` stage from
 `.pipelines/v2/templates/stages-test-engine.yml`. It runs on the existing `onnxruntime-Ubuntu2404-AMD-CPU` pool image
-without a custom container and uses standard CPU NuGet packages with a test-only Engine-capable GenAI pin.
-`FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS=ON` rejects packages that cannot compile the suite. The XML gate rejects
+without a custom container and uses the same stable ORT and GenAI packages as the SDK build. The XML gate rejects
 missing lifecycle cases, skipped/disabled/unexecuted tests, and failures. Do not bypass failures with skips or
-`continueOnError`. Its binaries are not published as SDK artifacts, and shipping/release dependency pins are unchanged.
-Generator-only builds can still exclude the suite; the unconditional Engine lane supplies the required coverage.
+`continueOnError`. Its binaries are not published as SDK artifacts.
 
 ## Debugging skips in CI
 

diff --git a/.pipelines/foundry-local-packaging.yml b/.pipelines/foundry-local-packaging.yml
@@ -53,17 +53,12 @@ parameters:
 
 variables:
 - group: FoundryLocal-ESRP-Signing
-# C++ SDK (sdk_v2/cpp) native dependency versions. Release builds use the
-# stable pins from sdk_v2/deps_versions.json; non-release CI validates the
-# selected GenAI nightly before it becomes a stable SDK default.
+# C++ SDK (sdk_v2/cpp) native dependency versions. Must match the stable pins
+# in sdk_v2/deps_versions.json.
 - name: cppOrtVersion
-  value: '1.28.0'
-- ${{ if eq(parameters.isRelease, true) }}:
-  - name: cppGenaiVersion
-    value: '0.15.2'
-- ${{ else }}:
-  - name: cppGenaiVersion
-    value: '0.16.0-dev1001400138'
+  value: '1.30.0'
+- name: cppGenaiVersion
+  value: '0.16.0'
 - name: cppWinmlVersion
   value: '2.1.70'
 - name: cppBuildConfig

diff --git a/.pipelines/v2/sdk_v2-pipeline-plan.md b/.pipelines/v2/sdk_v2-pipeline-plan.md
@@ -284,9 +284,8 @@ purposes:
 
 Versions are pipeline-level variables, currently:
 
-* `ortVersion`        `1.28.0`   (`Microsoft.ML.OnnxRuntime`)
-* `genaiVersion`      `0.15.2` for releases; selected ORT-Nightly version for non-release CI
-  (`Microsoft.ML.OnnxRuntimeGenAI.Foundry`)
+* `ortVersion`        `1.30.0`   (`Microsoft.ML.OnnxRuntime`)
+* `genaiVersion`      `0.16.0`   (`Microsoft.ML.OnnxRuntimeGenAI.Foundry`)
 * `winmlVersion`      `2.1.70`    (`Microsoft.Windows.AI.MachineLearning`, WinML 2.x reg-free)
 
 These must be kept in sync with the cmake defaults and with

diff --git a/.pipelines/v2/templates/stages-sdk-v2.yml b/.pipelines/v2/templates/stages-sdk-v2.yml
@@ -33,11 +33,12 @@ stages:
     genaiVersion: ${{ parameters.genaiVersion }}
     winmlVersion: ${{ parameters.winmlVersion }}
 
-# Test-only CUDA runtime; never replaces the native packaging artifacts.
+# Dedicated Engine lifecycle test stage.
 - template: stages-test-engine.yml
   parameters:
     buildConfig: ${{ parameters.buildConfig }}
     ortVersion: ${{ parameters.ortVersion }}
+    genaiVersion: ${{ parameters.genaiVersion }}
 
 # ── C# SDK ──
 - template: stages-cs.yml

diff --git a/.pipelines/v2/templates/stages-test-engine.yml b/.pipelines/v2/templates/stages-test-engine.yml
@@ -1,4 +1,4 @@
-# CPU paged-attention Engine gate, independent of the Generator/release packaging matrix.
+# CPU paged-attention Engine gate.
 # Model assets are checked in and staged by the normal C++ testdata build rule.
 parameters:
 - name: buildConfig
@@ -7,7 +7,6 @@ parameters:
   type: string
 - name: genaiVersion
   type: string
-  default: '0.16.0-dev1001407373'
 
 stages:
 - stage: cpp_test_engine
@@ -48,8 +47,7 @@ stages:
           --cmake_extra_defines \
             "GENAI_FETCH_URL=$NUGET_CACHE/genai.zip" \
             "ORT_GENAI_VERSION=$ENGINE_GENAI_VERSION" \
-            "ORT_FETCH_URL=$NUGET_CACHE/ort.zip" \
-            "FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS=ON"
+            "ORT_FETCH_URL=$NUGET_CACHE/ort.zip"
         build="$PWD/build/Linux/$BUILD_CONFIG"
         bin="$build/bin"
         export LD_LIBRARY_PATH="$bin:${LD_LIBRARY_PATH:-}"

diff --git a/sdk_v2/cpp/CMakeLists.txt b/sdk_v2/cpp/CMakeLists.txt
@@ -50,12 +50,6 @@ option(FOUNDRY_LOCAL_BUILD_TESTS "Build unit tests" ON)
 option(FOUNDRY_LOCAL_BUILD_EXAMPLES "Build example programs" ON)
 option(FOUNDRY_LOCAL_BUILD_SERVICE "Build web service support (requires oat++)" ON)
 option(FOUNDRY_LOCAL_ENABLE_ASAN "Enable AddressSanitizer + UndefinedBehaviorSanitizer (Linux only)" OFF)
-option(FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS
-       "Fail configuration unless the ORT GenAI dynamic Engine tests can be built"
-       OFF)
-if(FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS AND NOT FOUNDRY_LOCAL_BUILD_TESTS)
-    message(FATAL_ERROR "FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS requires FOUNDRY_LOCAL_BUILD_TESTS=ON")
-endif()
 
 # Optional 1DS ingestion token override. Override only via environment so it does
 # not appear in CMake cache files.
@@ -115,50 +109,6 @@ endif()
 # ORT and ORT GenAI — acquired via FetchContent from nuget.org.
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
 find_package(OnnxRuntimeGenAI REQUIRED)
-
-include(CheckCXXSourceCompiles)
-
-set(CMAKE_REQUIRED_LIBRARIES OnnxRuntimeGenAI::OnnxRuntimeGenAI)
-unset(FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE CACHE)
-check_cxx_source_compiles("
-#include <cstddef>
-#include <cstdint>
-#include <memory>
-
-#include <ort_genai.h>
-
-int main() {
-    auto* model = reinterpret_cast<OgaModel*>(1);
-    auto engine = OgaEngine::Create(*model);
-    auto events = engine->CreateEventBuffer(1);
-    auto request_options = OgaRequestOptions::Create();
-    auto request = engine->CreateRequest(request_options.get());
-    auto turn_options = request->CreateTurnOptions();
-    turn_options->SetMaxGeneratedTokens(static_cast<uint64_t>(1));
-    turn_options->SetDoSample(true);
-    turn_options->SetTemperature(1.0f);
-    turn_options->SetTopP(1.0f);
-    turn_options->SetTopK(1);
-    turn_options->SetSeed(static_cast<uint64_t>(1));
-    const char* values[] = {\"stop\"};
-    auto stop_strings = OgaStringArray::Create(values, 1);
-    turn_options->SetStopStrings(*stop_strings);
-    turn_options->SetGuidance(\"\", \"\");
-    const auto finish_reason = OgaFinishReason_StopString;
-    return events && request && turn_options && finish_reason == OgaFinishReason_Eos ? 0 : 1;
-}
-" FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE)
-unset(CMAKE_REQUIRED_LIBRARIES)
-
-if(FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE)
-    message(STATUS "ORT GenAI dynamic Engine API: enabled")
-else()
-    if(FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS)
-        message(FATAL_ERROR
-            "FOUNDRY_LOCAL_REQUIRE_DYNAMIC_ENGINE_TESTS requires an ORT GenAI package with the dynamic Engine API")
-    endif()
-    message(STATUS "ORT GenAI dynamic Engine API: unavailable; using Generator backend")
-endif()
 find_package(OnnxRuntime REQUIRED)
 
 # WinML EP Catalog — Windows-only, for hardware EP discovery and download. The
@@ -331,12 +281,10 @@ set(FOUNDRY_LOCAL_SOURCES
     ${FOUNDRY_LOCAL_INTERNAL_HEADERS}
 )
 
-if(FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE)
-    list(APPEND FOUNDRY_LOCAL_SOURCES
-        src/inferencing/generative/chat/onnx_chat_engine.cc
-        src/inferencing/generative/chat/onnx_engine_chat_stream.cc
-    )
-endif()
+list(APPEND FOUNDRY_LOCAL_SOURCES
+    src/inferencing/generative/chat/onnx_chat_engine.cc
+    src/inferencing/generative/chat/onnx_engine_chat_stream.cc
+)
 
 # 1DS bridge — always compiled for Foundry Local Core.
 list(APPEND FOUNDRY_LOCAL_SOURCES src/telemetry/one_ds_telemetry.cc)
@@ -412,7 +360,6 @@ function(foundry_local_configure_target TARGET LINK_SCOPE)
         ${TARGET}
         PRIVATE
             FOUNDRY_LOCAL_HAS_EP_BOOTSTRAPPERS=$<BOOL:${FOUNDRY_LOCAL_HAS_EP_BOOTSTRAPPERS}>
-            FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE=$<BOOL:${FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE}>
     )
 
     if(FOUNDRY_LOCAL_BUILD_SERVICE)

diff --git a/sdk_v2/cpp/src/ep_detection/cuda_ep_manifest.cc b/sdk_v2/cpp/src/ep_detection/cuda_ep_manifest.cc
@@ -32,7 +32,7 @@ EpBundleArtifact Archive(std::string id, std::string filename, std::string sha25
 
 EpBundleManifest WindowsX64Manifest() {
   return EpBundleManifest{
-      .bundle_id = "cuda-ep-win-x64-cuda-12.8.4-ort-1.28.0-genai-0.15.2-20260806-182620",
+      .bundle_id = "cuda-ep-win-x64-cuda-12.8.4-ort-1.30.0-genai-0.16.0-20260913-071149",
       .artifacts =
           {
               Archive(
@@ -66,13 +66,13 @@ EpBundleManifest WindowsX64Manifest() {
                        .sha256 = "49487537744256a3d4365c4792b03bf31130ad1faea0a13eafa219620941d837"},
                   }),
               Archive(
-                  "cuda-ep", "cuda-ep-bins-win-x64-20260806-182620.zip",
-                  "e62938987e848a0fbb3d215dfefaed40307d2446393909927ba0345eaaf3d263", 256 * kMiB,
+                  "cuda-ep", "cuda-ep-bins-win-x64-20260913-071149.zip",
+                  "5d1c38eb4058b6898d78ae5e7881ac07dda8ef37156be232f0f63846a1405f85", 256 * kMiB,
                   {
                       {.relative_path = "onnxruntime-genai-cuda.dll",
-                       .sha256 = "7894fb5efaad4a663e834f20b912b44cc383629b24ffe8bbc6382786a7326dbc"},
+                       .sha256 = "132d28f988ee8bc1ce050e8dcdda7036ab436616d827508af8e0a4f6ec21f2da"},
                       {.relative_path = "onnxruntime_providers_cuda.dll",
-                       .sha256 = "60f1aeef7ebe27f7e659cb88f597005ca5a5e75832b85dcef3eef02b9322df9a"},
+                       .sha256 = "7d014892b64d03092c01e99ae5786c6080c5239eeee30e6fb902fd2a800fa95e"},
                   }),
           },
       .provider_relative_path = "onnxruntime_providers_cuda.dll",
@@ -81,7 +81,7 @@ EpBundleManifest WindowsX64Manifest() {
 
 EpBundleManifest WindowsArm64Manifest() {
   return EpBundleManifest{
-      .bundle_id = "cuda-ep-win-arm64-cuda-13.4.1-ort-1.28.0-genai-0.15.2-20260806-182803",
+      .bundle_id = "cuda-ep-win-arm64-cuda-13.4.1-ort-1.30.0-genai-0.16.0-20260913-071350",
       .artifacts =
           {
               Archive(
@@ -115,13 +115,13 @@ EpBundleManifest WindowsArm64Manifest() {
                        .sha256 = "c9e0ec0e0a4e659393e15897ed1f6e5bac677e0c0fe7e12290f0386f19477b6b"},
                   }),
               Archive(
-                  "cuda-ep", "cuda-ep-bins-win-arm64-20260806-182803.zip",
-                  "212e670c61b3292d4a7d98f16fc2cf61f7b080604e0c145e81c39ec81e7b3259", 96 * kMiB,
+                  "cuda-ep", "cuda-ep-bins-win-arm64-20260913-071350.zip",
+                  "fc8f0a01daafedc82aa57072f023b11cf801dd94ed0a449ddc8ac4698e43e519", 96 * kMiB,
                   {
                       {.relative_path = "onnxruntime-genai-cuda.dll",
-                       .sha256 = "ab61145f4bc6284286e663586f634b973072d58ced20c497c7e5259f2ef3fc08"},
+                       .sha256 = "d27a02b8a83d0aaac47904d1baff4f7954aa25dcc130dba539fe9aafc53eab30"},
                       {.relative_path = "onnxruntime_providers_cuda.dll",
-                       .sha256 = "d92ffbd23a84f91b976baed9031de267efe1dc892d85c09d0979d25b89f5d1a0"},
+                       .sha256 = "82a3887c64791fc7131a0f705130f02f456952915ecea71372236656c59a47b2"},
                   }),
           },
       .provider_relative_path = "onnxruntime_providers_cuda.dll",
@@ -130,17 +130,17 @@ EpBundleManifest WindowsArm64Manifest() {
 
 EpBundleManifest LinuxX64Manifest() {
   return EpBundleManifest{
-      .bundle_id = "cuda-ep-linux-x64-ort-1.28.0-genai-0.15.2-20260806-182830",
+      .bundle_id = "cuda-ep-linux-x64-ort-1.30.0-genai-0.16.0-20260913-071417",
       .artifacts =
           {
               Archive(
-                  "cuda-ep", "cuda-ep-linux-x64-20260806-182830.zip",
-                  "abf347e7234d7434105efde12a2e0609fdd1d8828167b9873f4463926f1206e6", 448 * kMiB,
+                  "cuda-ep", "cuda-ep-linux-x64-20260913-071417.zip",
+                  "482e615e4c66b3f14a980af00e330e28fa5897a34161ec2972f75289685546be", 448 * kMiB,
                   {
                       {.relative_path = "libonnxruntime-genai-cuda.so",
-                       .sha256 = "d5300fc4413d9e74bd8dfceb5233fca6fcfa1d5ddc247081365fdb5f143091e6"},
+                       .sha256 = "3c95bb3b92f560718482dd3638388b6fd7374fb5ba6d9d5f4c4ca22c095e7681"},
                       {.relative_path = "libonnxruntime_providers_cuda.so",
-                       .sha256 = "b88d7b7f4b2e81d3eff41663fc70f4ae9e03dee9e2301cb53dc250e5a96d7f7a"},
+                       .sha256 = "8589ae31ed4941e72693a6a2e9dd8b8e721c888cfd9e6ffcb26da4e6e58ce720"},
                   }),
           },
       .provider_relative_path = "libonnxruntime_providers_cuda.so",

diff --git a/sdk_v2/cpp/src/inferencing/generative/chat/chat_session.cc b/sdk_v2/cpp/src/inferencing/generative/chat/chat_session.cc
@@ -8,9 +8,7 @@
 #include "inferencing/generative/chat/media_input.h"
 #include "inferencing/generative/chat/onnx_chat_engine.h"
 #include "inferencing/generative/chat/onnx_chat_generator.h"
-#if FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE
 #include "inferencing/generative/chat/onnx_engine_chat_stream.h"
-#endif
 #include "inferencing/generative/chat/reasoning_stream_splitter.h"
 #include "inferencing/generative/chat/stop_strings.h"
 #include "inferencing/generative/genai_model_instance.h"
@@ -61,12 +59,7 @@ std::unique_ptr<ChatGenerator> CreateTextChatGenerator(const std::vector<Transcr
                                                        const ToolCallContext& tool_ctx,
                                                        bool use_full_context) {
   if (model.GetGenAIConfig().GetChatBackendKind() != ChatBackendKind::kGenerator) {
-#if FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE
     return OnnxEngineChatStream::Create(messages, options, model, tool_ctx);
-#else
-    FL_THROW(FOUNDRY_LOCAL_ERROR_INTERNAL,
-             "model requires the ORT GenAI dynamic Engine API, but this build does not provide it");
-#endif
   }
 
   return OnnxChatGenerator::Create(messages, options, model, tool_ctx, use_full_context);

diff --git a/sdk_v2/cpp/src/inferencing/generative/genai_model_instance.cc b/sdk_v2/cpp/src/inferencing/generative/genai_model_instance.cc
@@ -78,18 +78,12 @@ GenAIModelInstance::GenAIModelInstance(std::string model_id,
   }
 
   if (genai_config_.GetChatBackendKind() == ChatBackendKind::kEngine) {
-#if FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE
     try {
       chat_engine_ = std::make_unique<OnnxChatEngine>(*this);
     } catch (const std::runtime_error& e) {
       FL_LOG_AND_THROW(logger, FOUNDRY_LOCAL_ERROR_INTERNAL,
                        "failed to create chat engine for model ", model_id_, ": ", e.what());
     }
-#else
-    FL_LOG_AND_THROW(logger, FOUNDRY_LOCAL_ERROR_INTERNAL,
-                     "model ", model_id_,
-                     " requires the ORT GenAI dynamic Engine API, but this build does not provide it");
-#endif
   }
 }
 

diff --git a/sdk_v2/cpp/src/inferencing/generative/genai_model_instance.h b/sdk_v2/cpp/src/inferencing/generative/genai_model_instance.h
@@ -62,11 +62,7 @@ class GenAIModelInstance {
   /// Access the underlying OGA objects.
   OgaModel& GetOgaModel();
   Preprocessor& GetPreprocessor();
-#if FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE
   OnnxChatEngine* GetChatEngine() { return chat_engine_.get(); }
-#else
-  OnnxChatEngine* GetChatEngine() { return nullptr; }
-#endif
 
   /// Get the last-activity timestamp.
   std::chrono::steady_clock::time_point LastActivity() const { return last_activity_; }
@@ -93,9 +89,7 @@ class GenAIModelInstance {
   ExecutionProvider ep_;
   std::unique_ptr<OgaModel> oga_model_;
   std::unique_ptr<Preprocessor> preprocessor_;
-#if FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE
   std::unique_ptr<OnnxChatEngine> chat_engine_;
-#endif
   TagInfo tag_info_;
   std::once_flag tag_info_init_flag_;
   std::chrono::steady_clock::time_point last_activity_;

diff --git a/sdk_v2/cpp/test/CMakeLists.txt b/sdk_v2/cpp/test/CMakeLists.txt
@@ -95,17 +95,11 @@ if(FOUNDRY_LOCAL_HAS_EP_BOOTSTRAPPERS)
     )
 endif()
 
-if(FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE)
-    target_sources(foundry_local_tests PRIVATE
-        internal_api/chat/dynamic_engine_chat_test.cc
-    )
-endif()
+target_sources(foundry_local_tests PRIVATE
+    internal_api/chat/dynamic_engine_chat_test.cc
+)
 
 target_compile_options(foundry_local_tests PRIVATE ${FOUNDRY_LOCAL_COMPILE_OPTIONS})
-target_compile_definitions(
-    foundry_local_tests
-    PRIVATE FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE=$<BOOL:${FOUNDRY_LOCAL_OGA_HAS_DYNAMIC_ENGINE}>
-)
 
 target_link_libraries(foundry_local_tests
     PRIVATE