docs: retain localized runtime release guidance

Signed-off-by: LauraGPT <[email protected]>

README_ja.md

@@ -152,6 +152,8 @@ CPU/エッジで Python なしのオフライン ASR が必要な場合は、lla
 
 **事前ビルド済みバイナリ：** [Releases](https://github.com/modelscope/FunASR/releases) · [v0.2.6](https://github.com/modelscope/FunASR/releases/tag/runtime-llamacpp-v0.2.6) · [Linux Vulkan tarball](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-linux-x64-vulkan.tar.gz) · [Windows Vulkan zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-vulkan.zip) · [Windows CUDA zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-cuda.zip) · [Windows Blackwell CUDA zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-cuda-blackwell.zip) · **ダウンロードとクイックスタート：** [funasr.com/deploy/llama-cpp](https://www.funasr.com/en/deploy/llama-cpp.html) · **GGUF モデル：** [Hugging Face](https://huggingface.co/FunAudioLLM) · **ドキュメントとベンチマーク：** [runtime/llama.cpp/](./runtime/llama.cpp/)
 
+Windows GPU では、[runtime-llamacpp-v0.2.6](https://github.com/modelscope/FunASR/releases/tag/runtime-llamacpp-v0.2.6) の `windows-x64-vulkan`、`windows-x64-cuda` または `windows-x64-cuda-blackwell` パッケージを選択してください。RTX 50 / Blackwell (`sm_120`) には専用の `windows-x64-cuda-blackwell` パッケージがあります。CI のアーカイブ検証は実機での Blackwell 推論を保証しません。詳細は [llama.cpp 配布ガイド](https://www.funasr.com/en/deploy/llama-cpp.html) を参照してください。
+
 [Colab quickstart →](./examples/colab/README_ja.md) · [OpenAI API example →](./examples/openai_api/README_ja.md) · [Client recipes →](./examples/openai_api/CLIENTS.md) · [Workflow recipes →](./examples/openai_api/WORKFLOWS.md) · [Postman collection →](./examples/openai_api/POSTMAN.md) · [OpenAPI spec →](./examples/openai_api/OPENAPI.md) · [Security guide →](./examples/openai_api/SECURITY.md) · [Deployment matrix →](./docs/deployment_matrix_ja.md) · [デプロイドキュメント →](./runtime/readme.md) · [Agent連携 →](https://modelscope.github.io/FunASR/agent.html)
 
 ---

README_ko.md

@@ -152,6 +152,8 @@ CPU/엣지에서 Python 없이 오프라인 ASR만 필요하다면 llama.cpp / G
 
 **사전 빌드 바이너리:** [Releases](https://github.com/modelscope/FunASR/releases) · [v0.2.6](https://github.com/modelscope/FunASR/releases/tag/runtime-llamacpp-v0.2.6) · [Linux Vulkan tarball](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-linux-x64-vulkan.tar.gz) · [Windows Vulkan zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-vulkan.zip) · [Windows CUDA zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-cuda.zip) · [Windows Blackwell CUDA zip](https://github.com/modelscope/FunASR/releases/download/runtime-llamacpp-v0.2.6/funasr-llamacpp-windows-x64-cuda-blackwell.zip) · **다운로드와 빠른 시작:** [funasr.com/deploy/llama-cpp](https://www.funasr.com/en/deploy/llama-cpp.html) · **GGUF 모델:** [Hugging Face](https://huggingface.co/FunAudioLLM) · **문서와 벤치마크:** [runtime/llama.cpp/](./runtime/llama.cpp/)
 
+Windows GPU에서는 [runtime-llamacpp-v0.2.6](https://github.com/modelscope/FunASR/releases/tag/runtime-llamacpp-v0.2.6)의 `windows-x64-vulkan`, `windows-x64-cuda` 또는 `windows-x64-cuda-blackwell` 패키지를 선택하세요. RTX 50 / Blackwell (`sm_120`)에는 전용 `windows-x64-cuda-blackwell` 패키지가 있습니다. CI 아카이브 검증은 실제 Blackwell 하드웨어에서의 추론을 보장하지 않습니다. 자세한 내용은 [llama.cpp 배포 가이드](https://www.funasr.com/en/deploy/llama-cpp.html)를 참조하세요.
+
 [Colab quickstart →](./examples/colab/README_ko.md) · [OpenAI API example →](./examples/openai_api/README_ko.md) · [Client recipes →](./examples/openai_api/CLIENTS.md) · [Workflow recipes →](./examples/openai_api/WORKFLOWS.md) · [Postman collection →](./examples/openai_api/POSTMAN.md) · [OpenAPI spec →](./examples/openai_api/OPENAPI.md) · [Security guide →](./examples/openai_api/SECURITY.md) · [Deployment matrix →](./docs/deployment_matrix_ko.md) · [배포 문서 →](./runtime/readme.md) · [Agent 연동 →](https://modelscope.github.io/FunASR/agent.html)
 
 ---

runtime/llama.cpp/tests/test_release_workflow.py

@@ -150,7 +150,7 @@ def test_readmes_surface_current_windows_vulkan_release():
     for name in ("README.md", "README_zh.md", "README_ja.md", "README_ko.md"):
         readme = (ROOT / name).read_text(encoding="utf-8")
 
-        assert "runtime-llamacpp-v0.1.9" in readme
+        assert "runtime-llamacpp-v0.2.6" in readme
         assert "windows-x64-vulkan" in readme