diff --git a/.ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 93%
rename from .ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
rename to .ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index 2569f0525..28f79955d 100644
--- a/.ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
+++ b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -16,8 +16,6 @@ cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
 - '12.9'
-cudnn:
-- '9'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -55,7 +55,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +67,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-64
 zip_keys:
diff --git a/.ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
new file mode 100644
index 000000000..166a18a83
--- /dev/null
+++ b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
@@ -0,0 +1,83 @@
+blas_impl:
+- generic
+c_compiler:
+- gcc
+c_compiler_version:
+- '14'
+c_stdlib:
+- sysroot
+c_stdlib_version:
+- '2.17'
+channel_sources:
+- conda-forge
+channel_targets:
+- conda-forge main
+cuda_compiler:
+- cuda-nvcc
+cuda_compiler_version:
+- None
+cxx_compiler:
+- gxx
+cxx_compiler_version:
+- '14'
+docker_image:
+- quay.io/condaforge/linux-anvil-x86_64:alma9
+fmt:
+- '12.1'
+github_actions_labels:
+- cirun-openstack-cpu-xlarge
+is_rc:
+- 'False'
+libabseil:
+- '20250512'
+libblas:
+- 3.9.* *netlib
+libcblas:
+- 3.9.* *netlib
+libcudnn_dev:
+- '9'
+liblapack:
+- 3.9.* *netlib
+libmagma_devel:
+- '2.9'
+libmagma_sparse:
+- '2.9'
+libprotobuf:
+- 6.31.1
+libtorch:
+- '2.9'
+mkl:
+- '2025'
+mkl_devel:
+- '2025'
+nccl:
+- '2'
+numpy:
+- '2'
+orc:
+- 2.2.2
+pin_run_as_build:
+  python:
+    min_pin: x.x
+    max_pin: x.x
+pybind11_abi:
+- '11'
+python:
+- 3.10.* *_cpython
+- 3.11.* *_cpython
+- 3.12.* *_cpython
+- 3.13.* *_cp313
+- 3.14.* *_cp314
+pytorch:
+- '2.9'
+target_platform:
+- linux-64
+zip_keys:
+- - c_compiler_version
+  - cxx_compiler_version
+  - c_stdlib_version
+  - cuda_compiler_version
+- - channel_targets
+  - is_rc
+zlib:
+- '1'
diff --git a/.ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 91%
rename from .ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
rename to .ci_support/linux_64_blas_implgenericc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index 72f926c61..99210045a 100644
--- a/.ci_support/linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
+++ b/.ci_support/linux_64_blas_implgenericc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -7,7 +7,7 @@ c_compiler_version:
 c_stdlib:
 - sysroot
 c_stdlib_version:
-- '2.17'
+- '2.28'
 channel_sources:
 - conda-forge
 channel_targets:
@@ -15,9 +15,7 @@ channel_targets:
 cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
-- None
-cudnn:
-- '9'
+- '13.0'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -55,7 +55,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +67,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-64
 zip_keys:
diff --git a/.ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml b/.ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 93%
rename from .ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
rename to .ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index e26700b68..f9451da26 100644
--- a/.ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
+++ b/.ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -16,8 +16,6 @@ cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
 - '12.9'
-cudnn:
-- '9'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -55,7 +55,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +67,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-64
 zip_keys:
diff --git a/.ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml b/.ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
new file mode 100644
index 000000000..a86135bba
--- /dev/null
+++ b/.ci_support/linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
@@ -0,0 +1,83 @@
+blas_impl:
+- mkl
+c_compiler:
+- gcc
+c_compiler_version:
+- '14'
+c_stdlib:
+- sysroot
+c_stdlib_version:
+- '2.17'
+channel_sources:
+- conda-forge
+channel_targets:
+- conda-forge main
+cuda_compiler:
+- cuda-nvcc
+cuda_compiler_version:
+- None
+cxx_compiler:
+- gxx
+cxx_compiler_version:
+- '14'
+docker_image:
+- quay.io/condaforge/linux-anvil-x86_64:alma9
+fmt:
+- '12.1'
+github_actions_labels:
+- cirun-openstack-cpu-xlarge
+is_rc:
+- 'False'
+libabseil:
+- '20250512'
+libblas:
+- 3.9.* *netlib
+libcblas:
+- 3.9.* *netlib
+libcudnn_dev:
+- '9'
+liblapack:
+- 3.9.* *netlib
+libmagma_devel:
+- '2.9'
+libmagma_sparse:
+- '2.9'
+libprotobuf:
+- 6.31.1
+libtorch:
+- '2.9'
+mkl:
+- '2025'
+mkl_devel:
+- '2025'
+nccl:
+- '2'
+numpy:
+- '2'
+orc:
+- 2.2.2
+pin_run_as_build:
+  python:
+    min_pin: x.x
+    max_pin: x.x
+pybind11_abi:
+- '11'
+python:
+- 3.10.* *_cpython
+- 3.11.* *_cpython
+- 3.12.* *_cpython
+- 3.13.* *_cp313
+- 3.14.* *_cp314
+pytorch:
+- '2.9'
+target_platform:
+- linux-64
+zip_keys:
+- - c_compiler_version
+  - cxx_compiler_version
+  - c_stdlib_version
+  - cuda_compiler_version
+- - channel_targets
+  - is_rc
+zlib:
+- '1'
diff --git a/.ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml b/.ci_support/linux_64_blas_implmklc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 91%
rename from .ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
rename to .ci_support/linux_64_blas_implmklc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index ba0126a43..e4de82f78 100644
--- a/.ci_support/linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
+++ b/.ci_support/linux_64_blas_implmklc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -7,7 +7,7 @@ c_compiler_version:
 c_stdlib:
 - sysroot
 c_stdlib_version:
-- '2.17'
+- '2.28'
 channel_sources:
 - conda-forge
 channel_targets:
@@ -15,9 +15,7 @@ channel_targets:
 cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
-- None
-cudnn:
-- '9'
+- '13.0'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -55,7 +55,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +67,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-64
 zip_keys:
diff --git a/.ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml b/.ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 93%
rename from .ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
rename to .ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index f25102f27..be9c523b4 100644
--- a/.ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse.yaml
+++ b/.ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -16,8 +16,6 @@ cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
 - '12.9'
-cudnn:
-- '9'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -53,7 +53,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -65,8 +65,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-aarch64
 zip_keys:
diff --git a/.ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml b/.ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
new file mode 100644
index 000000000..1cad89097
--- /dev/null
+++ b/.ci_support/linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse.yaml
@@ -0,0 +1,81 @@
+blas_impl:
+- generic
+c_compiler:
+- gcc
+c_compiler_version:
+- '13'
+c_stdlib:
+- sysroot
+c_stdlib_version:
+- '2.17'
+channel_sources:
+- conda-forge
+channel_targets:
+- conda-forge main
+cuda_compiler:
+- cuda-nvcc
+cuda_compiler_version:
+- None
+cxx_compiler:
+- gxx
+cxx_compiler_version:
+- '13'
+docker_image:
+- quay.io/condaforge/linux-anvil-x86_64:alma9
+fmt:
+- '12.1'
+github_actions_labels:
+- cirun-openstack-cpu-xlarge
+is_rc:
+- 'False'
+libabseil:
+- '20250512'
+libblas:
+- 3.9.* *netlib
+libcblas:
+- 3.9.* *netlib
+libcudnn_dev:
+- '9'
+liblapack:
+- 3.9.* *netlib
+libmagma_devel:
+- '2.9'
+libmagma_sparse:
+- '2.9'
+libprotobuf:
+- 6.31.1
+libtorch:
+- '2.9'
+mkl:
+- '2025'
+nccl:
+- '2'
+numpy:
+- '2'
+orc:
+- 2.2.2
+pin_run_as_build:
+  python:
+    min_pin: x.x
+    max_pin: x.x
+pybind11_abi:
+- '11'
+python:
+- 3.10.* *_cpython
+- 3.11.* *_cpython
+- 3.12.* *_cpython
+- 3.13.* *_cp313
+- 3.14.* *_cp314
+pytorch:
+- '2.9'
+target_platform:
+- linux-aarch64
+zip_keys:
+- - c_compiler_version
+  - cxx_compiler_version
+  - c_stdlib_version
+  - cuda_compiler_version
+- - channel_targets
+  - is_rc
+zlib:
+- '1'
diff --git a/.ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml b/.ci_support/linux_aarch64_c_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
similarity index 91%
rename from .ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
rename to .ci_support/linux_aarch64_c_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
index 2cd6684f4..85f26993c 100644
--- a/.ci_support/linux_aarch64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
+++ b/.ci_support/linux_aarch64_c_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse.yaml
@@ -7,7 +7,7 @@ c_compiler_version:
 c_stdlib:
 - sysroot
 c_stdlib_version:
-- '2.17'
+- '2.28'
 channel_sources:
 - conda-forge
 channel_targets:
@@ -15,9 +15,7 @@ channel_targets:
 cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
-- None
-cudnn:
-- '9'
+- '13.0'
 cxx_compiler:
 - gxx
 cxx_compiler_version:
@@ -36,15 +34,17 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
-magma:
 - '2.9'
 mkl:
 - '2025'
@@ -53,7 +53,7 @@ nccl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -65,8 +65,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - linux-aarch64
 zip_keys:
diff --git a/.ci_support/migrations/cuda130.yaml b/.ci_support/migrations/cuda130.yaml
new file mode 100644
index 000000000..e280eed28
--- /dev/null
+++ b/.ci_support/migrations/cuda130.yaml
@@ -0,0 +1,65 @@
+migrator_ts: 1755016036
+__migrator:
+  operation: key_add
+  migration_number:
+    1
+  build_number:
+    1
+  paused: false
+  use_local: true
+  override_cbc_keys:
+    - cuda_compiler_stub
+  check_solvable: false
+  primary_key: cuda_compiler_version
+  ordering:
+    cuda_compiler_version:
+      - 12.4
+      - 12.6
+      - 12.8
+      - None
+      - 12.9
+      - 13.0
+      # to allow manual opt-in for CUDA 11.8, see
+      # https://github.com/conda-forge/conda-forge-pinning-feedstock/pull/7472
+      # must be last due to how cuda_compiler ordering in that migrator works
+      - 11.8
+  wait_for_migrators:
+    - cuda129
+  commit_message: |
+    Upgrade to CUDA 13.0
+    
+    CUDA 13.0 requires architecture `sm_75` or higher, and renamed `sm_101` to
+    `sm_110`. To build for these, maintainers will need to modify their existing list of
+    specified architectures (e.g. `CMAKE_CUDA_ARCHITECTURES`, `TORCH_CUDA_ARCH_LIST`, etc.)
+    for their package.
+
+    Since CUDA 12.8, the conda-forge nvcc package now sets `CUDAARCHS` and
+    in its activation script to a string containing all
+    of the supported real architectures plus the virtual architecture of the
+    latest. Recipes for packages who use these variables to control their build
+    but do not want to build for all supported architectures will need to override
+    these variables in their build script.
+
+    ref: https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html#new-features
+
+    > [[!IMPORTANT]]
+    > Remember to update any CUDA 11/12 specific selector syntax in the recipe to include
+    > CUDA 13. For example `# [(cuda_compiler_version or "None").startswith("12")]`
+    > might be replaced with `# [cuda_compiler_version != "None"]`.
+
+cuda_compiler_version:         # [((linux and (x86_64 or aarch64)) or win64) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 13.0                       # [((linux and (x86_64 or aarch64)) or win64) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+
+c_stdlib_version:              # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 2.28                       # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+
+# Compiler pinnings deviate from global migrator in order to honor pinnings in local conda_build_config for aarch64
+c_compiler_version:            # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 14                         # [(linux and (x86_64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 13                         # [(linux and (aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+cxx_compiler_version:          # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 14                         # [(linux and (x86_64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 13                         # [(linux and (aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+
+fortran_compiler_version:      # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
+  - 14                         # [(linux and (x86_64 or aarch64)) and os.environ.get("CF_CUDA_ENABLED", "False") == "True"]
diff --git a/.ci_support/migrations/mkl2025.yaml b/.ci_support/migrations/mkl2025.yaml
deleted file mode 100644
index 9975d5713..000000000
--- a/.ci_support/migrations/mkl2025.yaml
+++ /dev/null
@@ -1,10 +0,0 @@
-__migrator:
-  build_number: 1
-  commit_message: Rebuild for mkl 2025
-  kind: version
-  migration_number: 1
-migrator_ts: 1731082955.2367358
-mkl:
-- '2025'
-mkl_devel:
-- '2025'
diff --git a/.ci_support/migrations/python314.yaml b/.ci_support/migrations/python314.yaml
new file mode 100644
index 000000000..36ec6b413
--- /dev/null
+++ b/.ci_support/migrations/python314.yaml
@@ -0,0 +1,43 @@
+# this is intentionally sorted before the 3.13t migrator, because that determines
+# the order of application of the migrators; otherwise we'd have to add values for
+# is_freethreading and is_abi3 keys here, since that migration extends the zip;
+migrator_ts: 1724712607
+__migrator:
+    commit_message: Rebuild for python 3.14
+    migration_number: 1
+    operation: key_add
+    primary_key: python
+    ordering:
+        python:
+            - 3.9.* *_cpython
+            - 3.10.* *_cpython
+            - 3.11.* *_cpython
+            - 3.12.* *_cpython
+            - 3.13.* *_cp313
+            - 3.13.* *_cp313t
+            - 3.14.* *_cp314  # new entry
+    paused: false
+    longterm: true
+    pr_limit: 5
+    max_solver_attempts: 3  # this will make the bot retry "not solvable" stuff 12 times
+    exclude:
+        # this shouldn't attempt to modify the python feedstocks
+        - python
+        - pypy3.6
+        - pypy-meta
+        - cross-python
+        - python_abi
+    exclude_pinned_pkgs: false
+    ignored_deps_per_node:
+        matplotlib:
+            - pyqt
+    additional_zip_keys:
+        - channel_sources
+
+python:
+- 3.14.* *_cp314
+# additional entries to add for zip_keys
+is_python_min:
+- false
+channel_sources:
+- conda-forge,conda-forge/label/python_rc
diff --git a/.ci_support/osx_64_blas_implgenericchannel_targetsconda-forge_mainis_rcFalse.yaml b/.ci_support/osx_64_blas_implgenericchannel_targetsconda-forge_mainis_rcFalse.yaml
index d7641c86c..d1731f6a8 100644
--- a/.ci_support/osx_64_blas_implgenericchannel_targetsconda-forge_mainis_rcFalse.yaml
+++ b/.ci_support/osx_64_blas_implgenericchannel_targetsconda-forge_mainis_rcFalse.yaml
@@ -36,14 +36,18 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
+- '2.9'
 llvm_openmp:
 - '19'
 macos_machine:
@@ -55,7 +59,7 @@ mkl_devel:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +71,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - osx-64
 zip_keys:
diff --git a/.ci_support/osx_64_blas_implmklchannel_targetsconda-forge_mainis_rcFalse.yaml b/.ci_support/osx_64_blas_implmklchannel_targetsconda-forge_mainis_rcFalse.yaml
index 61cdf0aa6..8af35f1a6 100644
--- a/.ci_support/osx_64_blas_implmklchannel_targetsconda-forge_mainis_rcFalse.yaml
+++ b/.ci_support/osx_64_blas_implmklchannel_targetsconda-forge_mainis_rcFalse.yaml
@@ -36,14 +36,18 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
+- '2.9'
 llvm_openmp:
 - '19'
 macos_machine:
@@ -55,7 +59,7 @@ mkl_devel:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -67,8 +71,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - osx-64
 zip_keys:
diff --git a/.ci_support/osx_arm64_channel_targetsconda-forge_mainis_rcFalse.yaml b/.ci_support/osx_arm64_channel_targetsconda-forge_mainis_rcFalse.yaml
index c8188ea68..ccb3ff9f0 100644
--- a/.ci_support/osx_arm64_channel_targetsconda-forge_mainis_rcFalse.yaml
+++ b/.ci_support/osx_arm64_channel_targetsconda-forge_mainis_rcFalse.yaml
@@ -36,14 +36,18 @@ libblas:
 - 3.9.* *netlib
 libcblas:
 - 3.9.* *netlib
+libcudnn_dev:
+- '9'
 liblapack:
 - 3.9.* *netlib
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
+- '2.9'
 llvm_openmp:
 - '19'
 macos_machine:
@@ -53,7 +57,7 @@ mkl:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -65,8 +69,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - osx-arm64
 zip_keys:
diff --git a/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version12.8is_rcFalse.yaml b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version12.8is_rcFalse.yaml
index 9d3fe548e..6d035c4db 100644
--- a/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version12.8is_rcFalse.yaml
+++ b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version12.8is_rcFalse.yaml
@@ -12,8 +12,6 @@ cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
 - '12.8'
-cudnn:
-- '9'
 cxx_compiler:
 - vs2022
 fmt:
@@ -24,12 +22,16 @@ is_rc:
 - 'False'
 libabseil:
 - '20250512'
+libcudnn_dev:
+- '9'
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
+- '2.9'
 magma:
 - '2.9'
 mkl:
@@ -39,7 +41,7 @@ mkl_devel:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -51,8 +53,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - win-64
 zip_keys:
diff --git a/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version13.0is_rcFalse.yaml b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version13.0is_rcFalse.yaml
new file mode 100644
index 000000000..112748fa4
--- /dev/null
+++ b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_version13.0is_rcFalse.yaml
@@ -0,0 +1,65 @@
+blas_impl:
+- mkl
+c_compiler:
+- vs2022
+c_stdlib:
+- vs
+channel_sources:
+- conda-forge
+channel_targets:
+- conda-forge main
+cuda_compiler:
+- cuda-nvcc
+cuda_compiler_version:
+- '13.0'
+cxx_compiler:
+- vs2022
+fmt:
+- '12.1'
+github_actions_labels:
+- cirun-azure-windows-4xlarge
+is_rc:
+- 'False'
+libabseil:
+- '20250512'
+libcudnn_dev:
+- '9'
+libmagma_devel:
+- '2.9'
+libmagma_sparse:
+- '2.9'
+libprotobuf:
+- 6.31.1
+libtorch:
+- '2.9'
+magma:
+- '2.9'
+mkl:
+- '2025'
+mkl_devel:
+- '2025'
+numpy:
+- '2'
+orc:
+- 2.2.2
+pin_run_as_build:
+  python:
+    min_pin: x.x
+    max_pin: x.x
+pybind11_abi:
+- '11'
+python:
+- 3.10.* *_cpython
+- 3.11.* *_cpython
+- 3.12.* *_cpython
+- 3.13.* *_cp313
+- 3.14.* *_cp314
+pytorch:
+- '2.9'
+target_platform:
+- win-64
+zip_keys:
+- - channel_targets
+  - is_rc
+zlib:
+- '1'
diff --git a/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
index f0dc0d666..544015555 100644
--- a/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
+++ b/.ci_support/win_64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse.yaml
@@ -12,8 +12,6 @@ cuda_compiler:
 - cuda-nvcc
 cuda_compiler_version:
 - None
-cudnn:
-- '9'
 cxx_compiler:
 - vs2022
 fmt:
@@ -24,12 +22,16 @@ is_rc:
 - 'False'
 libabseil:
 - '20250512'
+libcudnn_dev:
+- '9'
+libmagma_devel:
+- '2.9'
 libmagma_sparse:
 - '2.9'
 libprotobuf:
 - 6.31.1
 libtorch:
-- '2.7'
+- '2.9'
 magma:
 - '2.9'
 mkl:
@@ -39,7 +41,7 @@ mkl_devel:
 numpy:
 - '2'
 orc:
-- 2.2.1
+- 2.2.2
 pin_run_as_build:
   python:
     min_pin: x.x
@@ -51,8 +53,9 @@ python:
 - 3.11.* *_cpython
 - 3.12.* *_cpython
 - 3.13.* *_cp313
+- 3.14.* *_cp314
 pytorch:
-- '2.7'
+- '2.9'
 target_platform:
 - win-64
 zip_keys:
diff --git a/.github/workflows/conda-build.yml b/.github/workflows/conda-build.yml
index ea4e0228b..72bac5665 100644
--- a/.github/workflows/conda-build.yml
+++ b/.github/workflows/conda-build.yml
@@ -21,46 +21,64 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - CONFIG: linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse
-            SHORT_CONFIG: linux_64_blas_implgenericchannel_targets_h71a96c7d
+          - CONFIG: linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implgenericc_stdlib_versio_h404bdc8f
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implgenericchannel_targets_h71a96c7d', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implgenericc_stdlib_versio_h404bdc8f', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
-          - CONFIG: linux_64_blas_implgenericchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse
-            SHORT_CONFIG: linux_64_blas_implgenericchannel_targets_h17c608a0
+          - CONFIG: linux_64_blas_implgenericc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implgenericc_stdlib_versio_hf19b8591
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implgenericchannel_targets_h17c608a0', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-cpu-xlarge--${{ github.run_id }}-linux_64_blas_implgenericc_stdlib_versio_hf19b8591', 'linux', 'x64', 'self-hosted']
+            DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
+          - CONFIG: linux_64_blas_implgenericc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implgenericc_stdlib_versio_h139667b8
+            UPLOAD_PACKAGES: True
+            os: ubuntu
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implgenericc_stdlib_versio_h139667b8', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
-          - CONFIG: linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse
-            SHORT_CONFIG: linux_64_blas_implmklchannel_targetscond_h5b18f8bc
+          - CONFIG: linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implmklc_stdlib_version2.1_hcfd1fbb5
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implmklchannel_targetscond_h5b18f8bc', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implmklc_stdlib_version2.1_hcfd1fbb5', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
-          - CONFIG: linux_64_blas_implmklchannel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse
-            SHORT_CONFIG: linux_64_blas_implmklchannel_targetscond_h38f93959
+          - CONFIG: linux_64_blas_implmklc_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implmklc_stdlib_version2.1_h2c83be21
+            UPLOAD_PACKAGES: True
+            os: ubuntu
+            runs_on: ['cirun-openstack-cpu-xlarge--${{ github.run_id }}-linux_64_blas_implmklc_stdlib_version2.1_h2c83be21', 'linux', 'x64', 'self-hosted']
+            DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
+          - CONFIG: linux_64_blas_implmklc_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_64_blas_implmklc_stdlib_version2.2_ha4aa5d9c
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implmklchannel_targetscond_h38f93959', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_64_blas_implmklc_stdlib_version2.2_ha4aa5d9c', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
-          - CONFIG: linux_aarch64_channel_targetsconda-forge_maincuda_compiler_version12.9is_rcFalse
-            SHORT_CONFIG: linux_aarch64_channel_targetsconda-forge_heefc8d83
+          - CONFIG: linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_version12.9github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_aarch64_c_stdlib_version2.17channe_h7b39afa9
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_aarch64_channel_targetsconda-forge_heefc8d83', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_aarch64_c_stdlib_version2.17channe_h7b39afa9', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
-          - CONFIG: linux_aarch64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse
-            SHORT_CONFIG: linux_aarch64_channel_targetsconda-forge_h56c2c839
+          - CONFIG: linux_aarch64_c_stdlib_version2.17channel_targetsconda-forge_maincuda_compiler_versionNonegithub_actions_labelscirun-openstack-cpu-xlargeis_rcFalse
+            SHORT_CONFIG: linux_aarch64_c_stdlib_version2.17channe_hbde90b87
             UPLOAD_PACKAGES: True
             os: ubuntu
-            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_aarch64_channel_targetsconda-forge_h56c2c839', 'linux', 'x64', 'self-hosted']
+            runs_on: ['cirun-openstack-cpu-xlarge--${{ github.run_id }}-linux_aarch64_c_stdlib_version2.17channe_hbde90b87', 'linux', 'x64', 'self-hosted']
+            DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
+          - CONFIG: linux_aarch64_c_stdlib_version2.28channel_targetsconda-forge_maincuda_compiler_version13.0github_actions_labelscirun-openstack-gpu-2xlargeis_rcFalse
+            SHORT_CONFIG: linux_aarch64_c_stdlib_version2.28channe_h4c53fe26
+            UPLOAD_PACKAGES: True
+            os: ubuntu
+            runs_on: ['cirun-openstack-gpu-2xlarge--${{ github.run_id }}-linux_aarch64_c_stdlib_version2.28channe_h4c53fe26', 'linux', 'x64', 'self-hosted']
             DOCKER_IMAGE: quay.io/condaforge/linux-anvil-x86_64:alma9
             CONDA_FORGE_DOCKER_RUN_ARGS: "--gpus all"
           - CONFIG: osx_64_blas_implgenericchannel_targetsconda-forge_mainis_rcFalse
@@ -83,6 +101,11 @@ jobs:
             UPLOAD_PACKAGES: True
             os: windows
             runs_on: ['cirun-azure-windows-4xlarge--${{ github.run_id }}-win_64_channel_targetsconda-forge_maincu_hca575dce', 'windows', 'x64', 'self-hosted']
+          - CONFIG: win_64_channel_targetsconda-forge_maincuda_compiler_version13.0is_rcFalse
+            SHORT_CONFIG: win_64_channel_targetsconda-forge_maincu_h28fca69f
+            UPLOAD_PACKAGES: True
+            os: windows
+            runs_on: ['cirun-azure-windows-4xlarge--${{ github.run_id }}-win_64_channel_targetsconda-forge_maincu_h28fca69f', 'windows', 'x64', 'self-hosted']
           - CONFIG: win_64_channel_targetsconda-forge_maincuda_compiler_versionNoneis_rcFalse
             SHORT_CONFIG: win_64_channel_targetsconda-forge_maincu_hc68ac914
             UPLOAD_PACKAGES: True
diff --git a/recipe/bld.bat b/recipe/bld.bat
index 5124f7a9f..6d217730c 100644
--- a/recipe/bld.bat
+++ b/recipe/bld.bat
@@ -93,12 +93,19 @@ if not "%cuda_compiler_version%" == "None" (
     set USE_NCCL=0
     set USE_STATIC_NCCL=0
 
-    @REM set CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v%desired_cuda%
-    @REM set CUDA_BIN_PATH=%CUDA_PATH%\bin
-
-    set "TORCH_CUDA_ARCH_LIST=5.0;6.0;7.0;7.5;8.0;8.6;9.0;10.0;12.0+PTX"
     set "TORCH_NVCC_FLAGS=-Xfatbin -compress-all"
 
+    if "%cuda_compiler_version:~0,2%"=="12" (
+        set "TORCH_CUDA_ARCH_LIST=5.0;6.0;7.0;7.5;8.0;8.6;9.0;10.0;12.0+PTX"
+    ) else if "%cuda_compiler_version%" == "13.0" (
+        set "TORCH_CUDA_ARCH_LIST=7.5;8.0;8.6;9.0;10.0;11.0;12.0+PTX"
+        REM c.f. https://github.com/pytorch/pytorch/pull/161316
+        set "TORCH_NVCC_FLAGS=!TORCH_NVCC_FLAGS! -compress-mode=size"
+    ) else (
+        echo "unsupported cuda version. edit build_pytorch.bat"
+        exit /b 1
+    )
+
     set MAGMA_HOME=%LIBRARY_PREFIX%
     set "PATH=%CUDA_BIN_PATH%;%PATH%"
     set CUDNN_INCLUDE_DIR=%LIBRARY_PREFIX%\include
diff --git a/recipe/build.sh b/recipe/build.sh
index fde7fe80a..ca502d559 100644
--- a/recipe/build.sh
+++ b/recipe/build.sh
@@ -212,6 +212,8 @@ elif [[ ${cuda_compiler_version} != "None" ]]; then
     export CUDAToolkit_TARGET_DIR=${PREFIX}/targets/${CUDA_TARGET}
     sed -i -e "s,@CUDA_TARGET@,${CUDA_TARGET}," torch/_inductor/cpp_builder.py
 
+    export TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
+
     # Compatibility matrix for update: https://en.wikipedia.org/wiki/CUDA#GPUs_supported
     # Warning from pytorch v1.12.1: In the future we will require one to
     # explicitly pass TORCH_CUDA_ARCH_LIST to cmake instead of implicitly
@@ -227,11 +229,16 @@ elif [[ ${cuda_compiler_version} != "None" ]]; then
         12.[89])
             export TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6;9.0;10.0;12.0+PTX"
             ;;
+        13.0)
+            export TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6;9.0;10.0;11.0;12.0+PTX"
+            # c.f. https://github.com/pytorch/pytorch/pull/161316
+            export TORCH_NVCC_FLAGS="$TORCH_NVCC_FLAGS -compress-mode=size"
+            ;;
         *)
             echo "No CUDA architecture list exists for CUDA v${cuda_compiler_version}. See build.sh for information on adding one."
             exit 1
     esac
-    export TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
+
     export NCCL_ROOT_DIR=$PREFIX
     export NCCL_INCLUDE_DIR=$PREFIX/include
     export USE_SYSTEM_NCCL=1
diff --git a/recipe/conda_build_config.yaml b/recipe/conda_build_config.yaml
index 1ce6c8f1f..0b1161e40 100644
--- a/recipe/conda_build_config.yaml
+++ b/recipe/conda_build_config.yaml
@@ -27,6 +27,7 @@ blas_impl:
 
 # https://github.com/conda-forge/.cirun
 github_actions_labels:
+- cirun-openstack-cpu-xlarge    # [linux]
 - cirun-openstack-gpu-2xlarge   # [linux]
 - cirun-macos-m4-large          # [osx]
 - cirun-azure-windows-4xlarge   # [win]
diff --git a/recipe/meta.yaml b/recipe/meta.yaml
index 51af0db04..07d8b87bb 100644
--- a/recipe/meta.yaml
+++ b/recipe/meta.yaml
@@ -1,6 +1,6 @@
 # if you wish to build release candidate number X, append the version string with ".rcX"
 {% set version = "2.10.0" %}
-{% set build = 0 %}
+{% set build = 1 %}
 
 # Use a higher build number for the CUDA variant, to ensure that it's
 # preferred by conda's solver, and it's preferentially
@@ -16,7 +16,7 @@
 # see https://github.com/pytorch/pytorch/blame/v{{ version }}/.ci/docker/ci_commit_pins/triton.txt
 # pytorch and triton are released in tandem, see notes in their release process
 # https://github.com/pytorch/pytorch/blob/main/RELEASE.md#triton-dependency-for-the-release
-{% set triton = "3.5.1" %}
+{% set triton = "3.6.0" %}
 
 package:
   name: libtorch
@@ -54,6 +54,8 @@ source:
       # backport https://github.com/pytorch/pytorch/pull/159828
       - patches/0014-Attempt-to-fix-torch.backends.cudnn.rnn-import.patch
       - patches/0015-Use-Intel-LLVM-openmp.patch
+      # continued $PREFIX/include headers
+      - patches/0016-Fix-TorchConfig.cmake.in-to-account-for-different-in.patch
       - patches_submodules/tensorpipe/0001-switch-away-from-find_package-CUDA.patch
 
 build:
@@ -66,9 +68,20 @@ build:
 {% else %}
   skip: true  # [is_rc]
 {% endif %}
+  # CUDA builds on GPU agents, non-CUDA builds on CPU agents; c.f.
+  # https://github.com/conda-forge/conda-smithy/pull/2434; the skip could be avoided
+  # using https://github.com/conda-forge/conda-forge-pinning-feedstock/pull/6910
+  # (plus a mechanism similar to CF_CUDA_ENABLED, preliminarily called CF_SELF_HOSTED).
+  # condition needs to be on a single line for conda-build to catch used variables
+  {% if (cuda_compiler_version == "None" and "gpu" in github_actions_labels) or (cuda_compiler_version != "None" and "cpu" in github_actions_labels) %}
+  skip: true
+  {% endif %}
   string: cuda{{ cuda_compiler_version | replace('.', '') }}_{{ blas_impl }}_h{{ PKG_HASH }}_{{ build }}  # [cuda_compiler_version != "None"]
   string: cpu_{{ blas_impl }}_h{{ PKG_HASH }}_{{ build }}                                                 # [cuda_compiler_version == "None"]
   detect_binary_files_with_prefix: false
+  missing_dso_whitelist:
+    - '*/libcuda.so.*'    # [linux]
+    - '*/nvcuda.dll'      # [win]
   run_exports:
     - {{ pin_subpackage('libtorch', max_pin='x.x') }}
   ignore_run_exports_from:
@@ -106,27 +119,29 @@ requirements:
     - rsync     # [unix]
   host:
     # GPU requirements
-    - cudnn                           # [cuda_compiler_version != "None"]
-    - nccl                            # [cuda_compiler_version != "None" and linux]
-    - magma                           # [cuda_compiler_version != "None"]
-    - cuda-version {{ cuda_compiler_version }}  # [cuda_compiler_version != "None"]
-    - nvtx-c                          # [cuda_compiler_version != "None"]
     {% if cuda_compiler_version != "None" %}
-    - cuda-driver-dev                 # [linux]
     - cuda-cudart-dev
     - cuda-cupti-dev
+    - cuda-driver-dev       # [linux]
+    - cuda-nvml-dev
     - cuda-nvrtc-dev
     - cuda-nvtx-dev
-    - cuda-nvml-dev
     - cuda-profiler-api
+    - cuda-version {{ cuda_compiler_version }}
     - cusparselt
     - libcublas-dev
+    - libcudnn-dev
     - libcudss-dev
-    - libcufile-dev  # [linux]
+    - libcufile-dev         # [linux]
     - libcufft-dev
     - libcurand-dev
     - libcusolver-dev
     - libcusparse-dev
+    - libmagma-devel        # [cuda_compiler_version != "12.8"]
+    # cannot (yet) use libmagma-devel due to https://github.com/conda-forge/libmagma-feedstock/issues/32
+    - magma                 # [cuda_compiler_version == "12.8"]
+    - nccl                  # [linux]
+    - nvtx-c
     {% endif %}
     # other requirements
     - python 3.12
@@ -249,28 +264,29 @@ outputs:
       host:
         - {{ pin_subpackage('libtorch', exact=True) }}
         # GPU requirements
-        - cudnn                           # [cuda_compiler_version != "None"]
-        - nccl                            # [cuda_compiler_version != "None" and linux]
-        - cuda-version {{ cuda_compiler_version }}  # [cuda_compiler_version != "None"]
-        - nvtx-c                          # [cuda_compiler_version != "None"]
-        # cannot (yet) use libmagma-devel due to https://github.com/conda-forge/libmagma-feedstock/issues/32
-        - magma                           # [cuda_compiler_version != "None"]
         {% if cuda_compiler_version != "None" %}
-        - cuda-driver-dev                 # [linux]
         - cuda-cudart-dev
         - cuda-cupti-dev
+        - cuda-driver-dev       # [linux]
+        - cuda-nvml-dev
         - cuda-nvrtc-dev
         - cuda-nvtx-dev
-        - cuda-nvml-dev
         - cuda-profiler-api
+        - cuda-version {{ cuda_compiler_version }}
         - cusparselt
         - libcublas-dev
+        - libcudnn-dev
         - libcudss-dev
-        - libcufile-dev  # [linux]
+        - libcufile-dev         # [linux]
         - libcufft-dev
         - libcurand-dev
         - libcusolver-dev
         - libcusparse-dev
+        - libmagma-devel        # [cuda_compiler_version != "12.8"]
+        # cannot (yet) use libmagma-devel due to https://github.com/conda-forge/libmagma-feedstock/issues/32
+        - magma                 # [cuda_compiler_version == "12.8"]
+        - nccl                  # [linux]
+        - nvtx-c
         {% endif %}
         # other requirements
         - python
@@ -516,7 +532,11 @@ outputs:
         {% set jobs = "-n 1" %}     # [linux64 and cuda_compiler_version != "None"]
         # test only one python version on aarch because emulation is super-slow;
         # disable hypothesis because it randomly yields health check errors
+
+        # the opengpu server has a card with sm_70, an architecture dropped by CUDA 13.0
+        {% if (cuda_compiler_version or "0").split(".")[0]|int < 13 %}
         - pytest {{ jobs }} {{ tests }} -k "not ({{ skips }})" -m "not hypothesis" --durations=50 --timeout=1200 --disable-warnings    # [not aarch64 or py==312]
+        {% endif %}
 
         # regression test for https://github.com/conda-forge/pytorch-cpu-feedstock/issues/329, where we picked up
         # duplicate `.pyc` files due to newest py-ver (3.13) in the build environment not matching the one in host;
@@ -526,8 +546,16 @@ outputs:
 
         # test integrity of CMake metadata and ensure that THPLayoutType is visible as a symbol from libtorch_python
         - cd cmake_test
-        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON $CMAKE_ARGS .   # [unix]
-        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON %CMAKE_ARGS% .  # [win]
+        - mkdir build build2
+        - cd build
+        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON $CMAKE_ARGS ..   # [unix]
+        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON %CMAKE_ARGS% ..  # [win]
+        - cmake --build .                   # [unix]
+        - cmake --build . --config Release  # [win]
+        - cd ../build2
+        # regression test for https://github.com/conda-forge/pytorch-cpu-feedstock/issues/479
+        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON -DTorch_DIR=$SP_DIR/torch/share/cmake/Torch $CMAKE_ARGS ..    # [unix]
+        - cmake -GNinja -DCMAKE_CXX_STANDARD=17 -DWITH_TORCH_PYTHON=ON -DTorch_DIR=%SP_DIR%/torch/share/cmake/Torch %CMAKE_ARGS% ..  # [win]
         - cmake --build .                   # [unix]
         - cmake --build . --config Release  # [win]
 
diff --git a/recipe/patches/0016-Fix-TorchConfig.cmake.in-to-account-for-different-in.patch b/recipe/patches/0016-Fix-TorchConfig.cmake.in-to-account-for-different-in.patch
new file mode 100644
index 000000000..4ef5c3b90
--- /dev/null
+++ b/recipe/patches/0016-Fix-TorchConfig.cmake.in-to-account-for-different-in.patch
@@ -0,0 +1,42 @@
+From 1d2b8bdebd64b3a7027a6f14b19ba481a6f6342d Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Micha=C5=82=20G=C3=B3rny?= <mgorny@gentoo.org>
+Date: Tue, 27 Jan 2026 19:31:46 +0100
+Subject: [PATCH 16/16] Fix TorchConfig.cmake.in to account for different
+ include dirs
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+Update the include logic to detect whether it's dealing with the include
+directory in site-packages or the top-level include directory, and set
+the paths appropriately.  In the former case, we need to reference
+`torch/csrc/api/include` from the top-level include directory.
+
+Signed-off-by: Michał Górny <mgorny@quansight.com>
+---
+ cmake/TorchConfig.cmake.in | 7 ++++---
+ 1 file changed, 4 insertions(+), 3 deletions(-)
+
+diff --git a/cmake/TorchConfig.cmake.in b/cmake/TorchConfig.cmake.in
+index 471b7154ed0..22c879f6d56 100644
+--- a/cmake/TorchConfig.cmake.in
++++ b/cmake/TorchConfig.cmake.in
+@@ -53,14 +53,15 @@ else()
+ endif()
+ 
+ # Include directories.
+-if(EXISTS "${TORCH_INSTALL_PREFIX}/include")
++if(EXISTS "${TORCH_INSTALL_PREFIX}/include/torch/csrc/api/include")
++  # top-level include directory
+   set(TORCH_INCLUDE_DIRS
+-    ${TORCH_INSTALL_PREFIX}/include
+     ${TORCH_INSTALL_PREFIX}/include/torch/csrc/api/include)
+ else()
++  # site-packages include directory
+   set(TORCH_INCLUDE_DIRS
+     ${TORCH_INSTALL_PREFIX}/include
+-    ${TORCH_INSTALL_PREFIX}/include/torch/csrc/api/include)
++    ${TORCH_INSTALL_PREFIX}/../../../../include/torch/csrc/api/include)
+ endif()
+ 
+ # Library dependencies.