diff --git a/.github/workflows/build-wheel.yml b/.github/workflows/build-wheel.yml
index 14efc9f798b..e757228cc49 100644
--- a/.github/workflows/build-wheel.yml
+++ b/.github/workflows/build-wheel.yml
@@ -72,8 +72,13 @@ jobs:
           # but not historical blobs.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Register problem matchers
         # Compiler and Cython errors from the wheel builds become annotations
diff --git a/.github/workflows/ci-pixi-source-test.yml b/.github/workflows/ci-pixi-source-test.yml
index a09f2e2756f..1e10d2a5a42 100644
--- a/.github/workflows/ci-pixi-source-test.yml
+++ b/.github/workflows/ci-pixi-source-test.yml
@@ -88,8 +88,13 @@ jobs:
           # cuda.core's "cuda.bindings 12.x or 13.x must be installed" guard.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Setup pixi
         uses: ./.github/actions/setup-pixi
@@ -147,8 +152,13 @@ jobs:
           # cuda.core's "cuda.bindings 12.x or 13.x must be installed" guard.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Setup pixi
         uses: ./.github/actions/setup-pixi
@@ -215,8 +225,13 @@ jobs:
           # cuda.core's "cuda.bindings 12.x or 13.x must be installed" guard.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Setup proxy cache
         uses: nv-gha-runners/setup-proxy-cache@main
diff --git a/.github/workflows/coverage.yml b/.github/workflows/coverage.yml
index f75ac200b06..50d168db042 100644
--- a/.github/workflows/coverage.yml
+++ b/.github/workflows/coverage.yml
@@ -71,8 +71,13 @@ jobs:
           # but not historical blobs.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Fix workspace ownership
         run: |
@@ -238,8 +243,13 @@ jobs:
           # but not historical blobs.
           fetch-depth: 0
           filter: blob:none
-          # third_party/cccl provides the CCCL headers cuda.core compiles against.
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Set up Python
         uses: $/.github/actions/setup-hardened-python
diff --git a/.github/workflows/lychee.yml b/.github/workflows/lychee.yml
index ecb609ef395..f8f79822b64 100644
--- a/.github/workflows/lychee.yml
+++ b/.github/workflows/lychee.yml
@@ -63,8 +63,14 @@ jobs:
           # Keep SCM history/tags without downloading historical source blobs.
           filter: blob:none
           persist-credentials: false
-          # Rendered docs build cuda.core against the pinned CCCL headers.
-          submodules: ${{ matrix.kind == 'rendered' }}
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        if: ${{ matrix.kind == 'rendered' }}
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Set up Python
         id: python
diff --git a/.github/workflows/test-sdist-linux.yml b/.github/workflows/test-sdist-linux.yml
index a4dad6edbc5..542ba2ebc93 100644
--- a/.github/workflows/test-sdist-linux.yml
+++ b/.github/workflows/test-sdist-linux.yml
@@ -47,7 +47,13 @@ jobs:
           # the commit graph (`git describe`) but not historical blobs.
           fetch-depth: 0
           filter: blob:none
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Set up Python
         uses: ./.github/actions/setup-hardened-python
diff --git a/.github/workflows/test-sdist-windows.yml b/.github/workflows/test-sdist-windows.yml
index bffcf5d2ea9..0c0563126d1 100644
--- a/.github/workflows/test-sdist-windows.yml
+++ b/.github/workflows/test-sdist-windows.yml
@@ -56,7 +56,13 @@ jobs:
           # the commit graph (`git describe`) but not historical blobs.
           fetch-depth: 0
           filter: blob:none
-          submodules: true
+          # Fetch CCCL separately: checkout's filter does not apply to submodules.
+          submodules: false
+
+      - name: Checkout CCCL submodule
+        run: |
+          git submodule sync --recursive
+          git submodule update --init --force --depth=1 --filter=blob:none --recursive
 
       - name: Set up Python
         uses: ./.github/actions/setup-hardened-python
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index df487ccfdc1..b614fdcc1e0 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -69,10 +69,11 @@ ran and any validation gaps.
 including the Developer Mode registry check.** Complete that setup before
 running the clone command.
 
-From the directory where you want to create the checkout, run:
+With Git 2.36 or newer, run this from the directory where you want to
+create the checkout:
 
 ```console
-$ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules https://github.com/NVIDIA/cuda-python.git
+$ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules --filter=blob:none --also-filter-submodules https://github.com/NVIDIA/cuda-python.git
 $ cd cuda-python
 ```
 
@@ -81,6 +82,9 @@ This recipe sets up the repository for source builds and development:
 - `core.symlinks=true` materializes symlinks during checkout, including the
   build-critical `cuda_core/_build_shared.py`.
 - `--recurse-submodules` initializes the CCCL headers needed to build `cuda.core`.
+- `--filter=blob:none --also-filter-submodules` skips historical file
+  contents in the repository and CCCL. Git downloads those contents
+  when needed; commit history and release tags remain available.
 - `submodule.recurse=true` keeps initialized submodules in sync on `git pull`,
   `git checkout`, and `git switch`.
 - The clone retains history and release tags for correct package versions.
diff --git a/ci/README.md b/ci/README.md
index 2fcd5b4ab1d..af468a8c745 100644
--- a/ci/README.md
+++ b/ci/README.md
@@ -24,6 +24,25 @@ The pinned checkout action's
 [`README`](https://github.com/actions/checkout/blob/3d3c42e5aac5ba805825da76410c181273ba90b1/README.md#fetch-all-history-for-all-tags-and-branches)
 documents `fetch-depth: 0` for fetching all history, branches, and tags.
 
+## CCCL Submodule Checkout
+
+Jobs that build `cuda.core` fetch its pinned CCCL headers in a separate step:
+set `submodules: false` on `actions/checkout`, then run:
+
+```bash
+git submodule sync --recursive
+git submodule update --init --force --depth=1 --filter=blob:none --recursive
+```
+
+The checkout action's `filter` input applies only to the root repository, so
+letting it fetch submodules downloads CCCL's unfiltered history. The separate
+step limits submodule history to one commit and skips historical blobs, while
+checking out the exact revision pinned by the root repository. Syncing URLs
+and forcing checkout preserve the action's behavior on reused workspaces.
+Keep `fetch-depth: 0` and `filter: blob:none` on the root checkout: package
+versioning still needs its complete ancestry and release tags. See
+[issue #3351](https://github.com/NVIDIA/cuda-python/issues/3351).
+
 ## Documentation Link Checks
 
 Two pre-commit hooks check links with one pinned native wheel (`lychee-bin`,
diff --git a/cuda_core/docs/source/install.rst b/cuda_core/docs/source/install.rst
index 0b4374392b9..e2d74063140 100644
--- a/cuda_core/docs/source/install.rst
+++ b/cuda_core/docs/source/install.rst
@@ -130,7 +130,7 @@ recipe below. On native Windows outside WSL, complete its Windows setup
 
 .. code-block:: console
 
-   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules https://github.com/NVIDIA/cuda-python.git
+   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules --filter=blob:none --also-filter-submodules https://github.com/NVIDIA/cuda-python.git
    $ cd cuda-python/cuda_core
    $ pip install .
 
@@ -161,7 +161,7 @@ For example, to work on ``cuda-core`` against CUDA 13:
 
 .. code-block:: console
 
-   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules https://github.com/NVIDIA/cuda-python.git
+   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules --filter=blob:none --also-filter-submodules https://github.com/NVIDIA/cuda-python.git
    $ cd cuda-python/cuda_core
    $ uv venv
    $ source .venv/bin/activate   # On Windows: .venv\Scripts\activate
@@ -184,7 +184,7 @@ For example, to work against CUDA 13:
 
 .. code-block:: console
 
-   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules https://github.com/NVIDIA/cuda-python.git
+   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules --filter=blob:none --also-filter-submodules https://github.com/NVIDIA/cuda-python.git
    $ cd cuda-python
    $ pixi run -e cu13 test-core
 
diff --git a/cuda_pathfinder/docs/source/install.rst b/cuda_pathfinder/docs/source/install.rst
index 6c2acbdd69a..cf49c40891e 100644
--- a/cuda_pathfinder/docs/source/install.rst
+++ b/cuda_pathfinder/docs/source/install.rst
@@ -64,7 +64,7 @@ before cloning.
 
 .. code-block:: console
 
-   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules https://github.com/NVIDIA/cuda-python.git
+   $ git clone --config core.symlinks=true --config submodule.recurse=true --recurse-submodules --filter=blob:none --also-filter-submodules https://github.com/NVIDIA/cuda-python.git
    $ cd cuda-python/cuda_pathfinder
    $ pip install .