diff --git a/.github/workflows/build-openvino.yml b/.github/workflows/build-openvino.yml
index 8879a6af1..86aba456c 100644
--- a/.github/workflows/build-openvino.yml
+++ b/.github/workflows/build-openvino.yml
@@ -33,7 +33,7 @@ env:
   LLAMA_ARG_LOG_PREFIX: 1
   LLAMA_ARG_LOG_TIMESTAMPS: 1
   # TODO: fix failing tests on OpenVINO backend
-  CTEST_EXCLUDE: "test-llama-archs|^test-recurrent-state-|test-backend-ops|test-save-load-state"
+  CTEST_EXCLUDE: "test-llama-archs|^test-recurrent-state-|test-save-load-state"
 
 jobs:
   ubuntu-24-openvino:
diff --git a/.github/workflows/build-self-hosted.yml b/.github/workflows/build-self-hosted.yml
index fda4879e2..02a38466f 100644
--- a/.github/workflows/build-self-hosted.yml
+++ b/.github/workflows/build-self-hosted.yml
@@ -395,7 +395,11 @@ jobs:
       - name: Test
         id: ggml-ci
         run: |
-          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_HIGH_PERF=1 GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+          LLAMA_ARG_THREADS=$(nproc) \
+          GG_BUILD_HIGH_PERF=1 \
+          GG_BUILD_NO_BF16=1 \
+          GG_BUILD_EXTRA_TESTS_0=1 \
+          bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
 
   cpu-arm64-graviton4-kleidiai:
     runs-on: ah-ubuntu_22_04-c8g_8x
@@ -434,6 +438,8 @@ jobs:
       - name: Test
         id: ggml-ci
         run: |
+          LLAMA_ARG_THREADS=$(nproc) \
           GG_BUILD_KLEIDIAI=1 \
           GG_BUILD_EXTRA_TESTS_0=1 \
-          bash ./ci/run.sh ./tmp/results ./tmp/mnt
+          GG_BUILD_HIGH_PERF=1 \
+          bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
diff --git a/.github/workflows/build-vulkan.yml b/.github/workflows/build-vulkan.yml
index 9de52e990..21d2a7735 100644
--- a/.github/workflows/build-vulkan.yml
+++ b/.github/workflows/build-vulkan.yml
@@ -164,9 +164,7 @@ jobs:
           export GGML_VK_VISIBLE_DEVICES=0
           export GGML_VK_DISABLE_F16=1
           export GGML_VK_DISABLE_COOPMAT=1
-          # This is using llvmpipe and runs slower than other backends
-          # test-backend-ops is too slow on llvmpipe, skip it
-          ctest -L main -E test-backend-ops --verbose --timeout 900
+          ctest -L main --verbose --timeout 900
 
   windows:
     runs-on: windows-2025
diff --git a/.github/workflows/build-webgpu.yml b/.github/workflows/build-webgpu.yml
index e624e3ba8..8277abcc4 100644
--- a/.github/workflows/build-webgpu.yml
+++ b/.github/workflows/build-webgpu.yml
@@ -190,6 +190,4 @@ jobs:
         id: cmake_test
         run: |
           cd build
-          # This is using llvmpipe and runs slower than other backends
-          # test-backend-ops is too slow on llvmpipe, skip it
-          ctest -L main -E test-backend-ops --verbose --timeout 900
+          ctest -L main --verbose --timeout 900
diff --git a/ci/run.sh b/ci/run.sh
index 294cbe57b..1ceb19fd5 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -190,7 +190,7 @@ if [ ! -z ${GG_BUILD_OPENVINO} ]; then
     CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_OPENVINO=ON"
 
     # TODO: fix failing tests on OpenVINO backend
-    CTEST_EXTRA="-E test-llama-archs|^test-recurrent-state-|test-backend-ops|test-save-load-state"
+    CTEST_EXTRA="-E test-llama-archs|^test-recurrent-state-|test-save-load-state"
 fi
 
 ## helpers
@@ -250,7 +250,7 @@ function gg_run_ctest_debug {
     (cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Debug ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
     (time cmake --build . --config Debug -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
 
-    (time ctest -C Debug --output-on-failure -L main -E "test-opt|test-backend-ops|test-llama-archs" ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
+    (time ctest -C Debug --output-on-failure -L main -E "test-opt|test-llama-archs" ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
 
     set +e
 }
@@ -768,25 +768,43 @@ function gg_check_build_requirements {
     fi
 }
 
-function gg_run_test_backend_ops_cpu {
+function gg_run_test_backend_ops {
     cd ${SRC}
 
     cd build-ci-release
 
     set -e
 
-    (time ./bin/test-backend-ops -b CPU ) 2>&1 | tee -a $OUT/${ci}-test-backend-ops-cpu.log
+    local args_extra="-j $(nproc)"
+
+    # TODO: fix multi-threaded for ROCm
+    #       https://github.com/ggml-org/llama.cpp/actions/runs/34576278519/job/103297889044?pr=28740#step:3:4865
+    if [ ! -z ${GG_BUILD_ROCM} ]; then
+        args_extra=""
+    fi
+
+    # TODO: MoltenVK bug?
+    #       https://github.com/ggml-org/llama.cpp/actions/runs/34611260059/job/103302413736?pr=28740#step:3:5897
+    if [ ! -z "${GG_BUILD_VULKAN}" ] && [ "$(uname -s)" = "Darwin" ]; then
+        args_extra=""
+    fi
+
+    if [ ! -z ${GG_BUILD_HIGH_PERF} ]; then
+        (time ./bin/test-backend-ops ${args_extra} -b CPU) 2>&1 | tee -a $OUT/${ci}-test-backend-ops.log
+    else
+        (time ./bin/test-backend-ops ${args_extra}       ) 2>&1 | tee -a $OUT/${ci}-test-backend-ops.log
+    fi
 
     set +e
 }
 
-function gg_sum_test_backend_ops_cpu {
+function gg_sum_test_backend_ops {
     gg_printf '### %s\n\n' "${ci}"
 
-    gg_printf 'Runs test-backend-ops for CPU backend\n'
+    gg_printf 'Runs test-backend-ops\n'
     gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
     gg_printf '```\n'
-    gg_printf '%s\n' "$(cat $OUT/${ci}-test-backend-ops-cpu.log)"
+    gg_printf '%s\n' "$(cat $OUT/${ci}-test-backend-ops.log)"
     gg_printf '```\n'
     gg_printf '\n'
 }
@@ -819,13 +837,11 @@ ret=0
 test $ret -eq 0 && gg_run ctest_debug
 test $ret -eq 0 && gg_run ctest_release
 
+test $ret -eq 0 && gg_run test_backend_ops
+
 test $ret -eq 0 && gg_run test_llama_archs_models
 test $ret -eq 0 && gg_run test_llama_archs_tensor_split
 
-if [ ! -z ${GG_BUILD_HIGH_PERF} ]; then
-    test $ret -eq 0 && gg_run test_backend_ops_cpu
-fi
-
 if [ -z ${GG_BUILD_LOW_PERF} ]; then
     test $ret -eq 0 && gg_run embd_bge_small
     test $ret -eq 0 && gg_run rerank_tiny
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 0c4e4d5a9..b3559a173 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -10,7 +10,7 @@ function(llama_build source)
     endif()
 
     add_executable(${TEST_TARGET} ${TEST_SOURCES})
-    target_link_libraries(${TEST_TARGET} PRIVATE llama llama-common)
+    target_link_libraries(${TEST_TARGET} PRIVATE llama llama-common) # TODO: [TAG_TESTS_LLAMA_LINK]
     if (LLAMA_TESTS_INSTALL)
         install(TARGETS ${TEST_TARGET} RUNTIME)
     endif()
@@ -310,15 +310,9 @@ if (NOT LLAMA_SANITIZE_ADDRESS AND NOT GGML_SCHED_NO_REALLOC)
   # TODO: repair known memory leaks
   llama_build_and_test(test-opt.cpp)
 endif()
-llama_build_and_test(test-backend-ops.cpp)
 
-# the tensor API kernels come from a separate metallib - check they produce correct results
-# ref: https://github.com/ggml-org/llama.cpp/issues/27473
-if (GGML_METAL AND NOT GGML_METAL_EMBED_LIBRARY)
-    llama_test(test-backend-ops NAME test-backend-ops-metallib-tensor
-               ARGS test -b MTL0 -o MUL_MAT -p type_a=q6_K)
-    set_tests_properties(test-backend-ops-metallib-tensor PROPERTIES ENVIRONMENT GGML_METAL_TENSOR_ENABLE=1)
-endif()
+# TODO: make this test (and others) not link `libllama` as it is not needed [TAG_TESTS_LLAMA_LINK]
+llama_build(test-backend-ops.cpp)
 
 llama_build_and_test(test-model-load-cancel.cpp LABEL "model")
 llama_build_and_test(test-autorelease.cpp       LABEL "model")