diff --git a/.github/backend-matrix.yml b/.github/backend-matrix.yml
index 2096d8d4f..2ce1b8c27 100644
--- a/.github/backend-matrix.yml
+++ b/.github/backend-matrix.yml
@@ -3167,6 +3167,14 @@ include:
   # No ROCm entry: upstream has no HIP configuration. No CUDA arm64 or L4T
   # entry: upstream documents and validates CUDA on x86 only. Darwin/Metal
   # lands with scripts/build/audio-cpp-darwin.sh, which does not exist yet.
+  #
+  # No vulkan entry either, though Dockerfile.audio-cpp and the backend Makefile
+  # both handle BUILD_TYPE=vulkan for local builds. Every other vulkan backend
+  # gets its Mesa ICD drivers from .docker/install-base-deps.sh, which installs
+  # mesa-vulkan-drivers so package-gpu-libs.sh can bundle them; this Dockerfile
+  # calls neither, so the image would ship a Vulkan loader that finds no GPU. No
+  # CI job runs a vulkan image against real hardware, so it would pass green and
+  # fail in users' hands. The entry comes back once the ICD question is settled.
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -3195,6 +3203,11 @@ include:
     dockerfile: "./backend/Dockerfile.audio-cpp"
     context: "./"
     ubuntu-version: '2404'
+  # cuda-major-version is forwarded into the build (Dockerfile.audio-cpp -> the
+  # backend Makefile) and picks the CMAKE_CUDA_ARCHITECTURES list, which upstream
+  # otherwise sets to `native` and no CI runner can enumerate. cuda-minor-version
+  # and the base-image tag encode the same toolkit and must move together;
+  # nothing checks that for you.
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "8"
@@ -3221,22 +3234,6 @@ include:
     dockerfile: "./backend/Dockerfile.audio-cpp"
     context: "./"
     ubuntu-version: '2404'
-  # Vulkan: ubuntu:24.04 plus apt's libvulkan-dev + glslc, installed by
-  # Dockerfile.audio-cpp when BUILD_TYPE=vulkan. arm64 vulkan is a one-line add
-  # once amd64 is proven in CI.
-  - build-type: 'vulkan'
-    cuda-major-version: ""
-    cuda-minor-version: ""
-    platforms: 'linux/amd64'
-    tag-latest: 'auto'
-    tag-suffix: '-gpu-vulkan-audio-cpp'
-    runs-on: 'ubuntu-latest'
-    base-image: "ubuntu:24.04"
-    skip-drivers: 'false'
-    backend: "audio-cpp"
-    dockerfile: "./backend/Dockerfile.audio-cpp"
-    context: "./"
-    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
diff --git a/backend/Dockerfile.audio-cpp b/backend/Dockerfile.audio-cpp
index 1172b6eba..99641076b 100644
--- a/backend/Dockerfile.audio-cpp
+++ b/backend/Dockerfile.audio-cpp
@@ -51,8 +51,14 @@ ARG TARGETARCH
 ARG TARGETVARIANT
 ARG APT_MIRROR
 ARG APT_PORTS_MIRROR
+# Selects the CUDA architecture list in backend/cpp/audio-cpp/Makefile. It has
+# to be forwarded: upstream compiles engine_runtime for `native` when
+# CMAKE_CUDA_ARCHITECTURES is unset, and no CI runner has a GPU to enumerate.
+# The value is the same cuda-major-version the matrix entry declares.
+ARG CUDA_MAJOR_VERSION
 
 ENV BUILD_TYPE=${BUILD_TYPE} \
+    CUDA_MAJOR_VERSION=${CUDA_MAJOR_VERSION} \
     APT_MIRROR=${APT_MIRROR} \
     APT_PORTS_MIRROR=${APT_PORTS_MIRROR} \
     DEBIAN_FRONTEND=noninteractive \
@@ -83,7 +89,8 @@ RUN --mount=type=bind,source=.docker/apt-mirror.sh,target=/usr/local/sbin/apt-mi
 COPY . /LocalAI
 
 RUN --mount=type=cache,target=/root/.ccache,id=audio-cpp-ccache-${TARGETARCH}-${BUILD_TYPE},sharing=locked \
-    make -C /LocalAI/backend/cpp/audio-cpp BUILD_TYPE=${BUILD_TYPE} NATIVE=false grpc-server package
+    make -C /LocalAI/backend/cpp/audio-cpp BUILD_TYPE=${BUILD_TYPE} \
+        CUDA_MAJOR_VERSION=${CUDA_MAJOR_VERSION} NATIVE=false grpc-server package
 
 # The package directory is the whole image: run.sh, grpc-server, the dlopened
 # ggml CPU variants, the bundled loader and its library closure, and the
diff --git a/backend/cpp/audio-cpp/Makefile b/backend/cpp/audio-cpp/Makefile
index c32d34a2f..d6e7e28aa 100644
--- a/backend/cpp/audio-cpp/Makefile
+++ b/backend/cpp/audio-cpp/Makefile
@@ -26,8 +26,39 @@ UNAME_S := $(shell uname -s)
 # safetensors model tree still resolves its family spec.
 CMAKE_ARGS ?= -DCMAKE_BUILD_TYPE=Release -DAUDIOCPP_DEPLOYMENT_BUILD=ON
 
+# CMAKE_CUDA_ARCHITECTURES must be set explicitly for a cublas build, and this
+# is not a tuning knob: upstream's CMakeLists sets CUDA_ARCHITECTURES to
+# `native` on the engine_runtime target whenever the root-scope variable is
+# unset (audio.cpp/CMakeLists.txt, the `if (CMAKE_CUDA_ARCHITECTURES)` branch
+# next to the istft/torch_random .cu sources), and docs/build/linux.md says so
+# outright: "Leave CMAKE_CUDA_ARCHITECTURES unset to build for the GPUs present
+# at build time (native)". No CI runner has a GPU, so `native` has nothing to
+# enumerate. ggml's own default (external/ggml/src/ggml-cuda/CMakeLists.txt)
+# does not rescue this: it list(APPEND)s in the ggml subdirectory scope, which
+# never reaches the root scope where the engine_runtime property is decided.
+#
+# The values below are ggml's list for the matching toolkit, copied rather than
+# invented, so the two targets compile for exactly the same set:
+#   - CUDA 13 drops the Maxwell/Pascal/Volta virtual archs (50/61/70).
+#   - 121a-real needs CUDA >= 12.9, so the CUDA 12 list (built against 12.8)
+#     stops at 120a-real.
+#   - `a`-suffixed archs are used rather than ggml's rejected 120f-virtual: the
+#     `f` suffix needs CMake >= 3.31.8, and Ubuntu Noble ships 3.28.3. The
+#     3.28 validator (Modules/Internal/CMakeCUDAArchitecturesValidate.cmake)
+#     accepts `[0-9]+a?(-real|-virtual)?`.
+#
+# Setting it here also pins ggml's copy, since its default is guarded by
+# `if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)`. CUDA_MAJOR_VERSION is the CI
+# build-arg, forwarded by Dockerfile.audio-cpp.
+CUDA_MAJOR_VERSION ?=
+ifeq ($(CUDA_MAJOR_VERSION),13)
+    CUDA_ARCHITECTURES ?= 75-virtual;80-virtual;86-real;89-real;120a-real;121a-real
+else
+    CUDA_ARCHITECTURES ?= 50-virtual;61-virtual;70-virtual;75-virtual;80-virtual;86-real;89-real;120a-real
+endif
+
 ifeq ($(BUILD_TYPE),cublas)
-    CMAKE_ARGS += -DENGINE_ENABLE_CUDA=ON
+    CMAKE_ARGS += -DENGINE_ENABLE_CUDA=ON "-DCMAKE_CUDA_ARCHITECTURES=$(CUDA_ARCHITECTURES)"
 else ifeq ($(BUILD_TYPE),vulkan)
     CMAKE_ARGS += -DENGINE_ENABLE_VULKAN=ON
 else ifeq ($(UNAME_S),Darwin)
diff --git a/backend/index.yaml b/backend/index.yaml
index 1572af117..429b55dec 100644
--- a/backend/index.yaml
+++ b/backend/index.yaml
@@ -148,16 +148,16 @@
     - audio-transcription
     - CPU
     - CUDA
-    - Vulkan
-  # No metal/metal-darwin-arm64 key: there is no Darwin matrix entry for
-  # audio-cpp yet, and pointing a capability at a tag CI never builds hands the
-  # user a pull failure instead of the CPU fallback below.
+  # No metal/metal-darwin-arm64 key (no Darwin matrix entry yet) and no vulkan
+  # key (the vulkan image would carry a Vulkan loader with no Mesa ICD; see the
+  # audio-cpp block in .github/backend-matrix.yml). Pointing a capability at a
+  # tag CI never builds hands the user a pull failure; leaving it out lets
+  # SystemState.Capability() fall through to `default` and the CPU image.
   capabilities:
     default: "cpu-audio-cpp"
     nvidia: "cuda12-audio-cpp"
     nvidia-cuda-12: "cuda12-audio-cpp"
     nvidia-cuda-13: "cuda13-audio-cpp"
-    vulkan: "vulkan-audio-cpp"
 - &whispercpp
   name: "whisper"
   alias: "whisper"
@@ -2037,7 +2037,6 @@
     nvidia: "cuda12-audio-cpp-development"
     nvidia-cuda-12: "cuda12-audio-cpp-development"
     nvidia-cuda-13: "cuda13-audio-cpp-development"
-    vulkan: "vulkan-audio-cpp-development"
 - !!merge <<: *stablediffusionggml
   name: "stablediffusion-ggml-development"
   capabilities:
@@ -6896,13 +6895,3 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-audio-cpp"
   mirrors:
     - localai/localai-backends:master-gpu-nvidia-cuda-13-audio-cpp
-- !!merge <<: *audiocpp
-  name: "vulkan-audio-cpp"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-audio-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-vulkan-audio-cpp
-- !!merge <<: *audiocpp
-  name: "vulkan-audio-cpp-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-audio-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-vulkan-audio-cpp
diff --git a/scripts/lib/backend-filter.mjs b/scripts/lib/backend-filter.mjs
index e9dee556a..c14a7c243 100644
--- a/scripts/lib/backend-filter.mjs
+++ b/scripts/lib/backend-filter.mjs
@@ -473,7 +473,7 @@ export const BACKEND_MATRIX_FILE = ".github/backend-matrix.yml";
 
 // Identity of a matrix entry across revisions. tag-suffix names the image;
 // per-arch legs of the same image are distinguished by platform-tag. Verified
-// unique across all 417 Linux and 56 Darwin entries.
+// unique across all 432 Linux and 57 Darwin entries.
 export function matrixEntryKey(item) {
   return JSON.stringify([item["tag-suffix"] || "", item["platform-tag"] || ""]);
 }