diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index bda02e9..d449dec 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -18,12 +18,9 @@ jobs:
       envs: ${{ steps.get-envs.outputs.envs }}
       pythons: ${{ steps.get-pythons.outputs.pythons }}
     steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          filter: blob:none
-      - name: Install uv
-        uses: astral-sh/setup-uv@v5
+      - uses: actions/checkout@v5
+        with: { fetch-depth: 0, filter: "blob:none" }
+      - uses: astral-sh/setup-uv@v7
       - name: Get test environments
         id: get-envs
         run: |
@@ -53,14 +50,10 @@ jobs:
               python: "3.13"
             os: macos-latest
     steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          filter: blob:none
-      - uses: astral-sh/setup-uv@v5
+      - uses: actions/checkout@v5
+        with: { fetch-depth: 0, filter: "blob:none" }
+      - uses: astral-sh/setup-uv@v7
         with:
-          enable-cache: true
-          cache-dependency-glob: pyproject.toml
           python-version: ${{ matrix.env.python }}
       - name: create environment
         run: uvx hatch env create ${{ matrix.env.name }}
@@ -80,33 +73,27 @@ jobs:
     name: CPU Benchmarks
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
+        with: { fetch-depth: 0, filter: "blob:none" }
       - uses: actions/setup-python@v5
         with:
           python-version: '3.13'
-      - uses: astral-sh/setup-uv@v5
-        with:
-          enable-cache: true
-          cache-dependency-glob: pyproject.toml
+      - uses: astral-sh/setup-uv@v7
       - run: uv pip install --system -e .[test,full]
       - uses: CodSpeedHQ/action@v3
         with:
-          run: pytest -m benchmark --codspeed
+          run: pytest -m benchmark --codspeed -n auto
           token: ${{ secrets.CODSPEED_TOKEN }}
   import:
     name: Import Tests
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          filter: blob:none
+      - uses: actions/checkout@v5
+        with: { fetch-depth: 0, filter: "blob:none" }
       - uses: actions/setup-python@v5
         with:
           python-version: '3.13'
-      - uses: astral-sh/setup-uv@v5
-        with:
-          cache-dependency-glob: pyproject.toml
+      - uses: astral-sh/setup-uv@v7
       - run: uv pip install --system -e .
       - run: python -c 'import fast_array_utils as fau; print(fau.__all__)'
       - run: uv pip install --system -e .[testing]
@@ -121,10 +108,8 @@ jobs:
     env:
       SKIP: no-commit-to-branch  # this CI runs on the main branch
     steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          filter: blob:none
+      - uses: actions/checkout@v5
+        with: { fetch-depth: 0, filter: "blob:none" }
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
diff --git a/pyproject.toml b/pyproject.toml
index a110131..edc5f27 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -50,6 +50,7 @@ optional-dependencies.test-min = [
   "pytest",
   "pytest-codspeed",
   "pytest-doctestplus",
+  "pytest-xdist",
 ]
 optional-dependencies.testing = [ "packaging" ]
 urls.'Documentation' = "https://icb-fast-array-utils.readthedocs-hosted.com/"
diff --git a/tests/test_stats.py b/tests/test_stats.py
index 984b460..54105b7 100644
--- a/tests/test_stats.py
+++ b/tests/test_stats.py
@@ -371,8 +371,10 @@ def test_stats_benchmark(
     axis: Literal[0, 1] | None,
     dtype: type[np.float32 | np.float64],
 ) -> None:
-    shape = (10_000, 10_000) if "sparse" in array_type.mod else (1000, 1000)
-    arr = array_type.random(shape, dtype=dtype)
+    # test with 10M elements will take 20ms for the fastest functions
+    n_elems, density = 10_000_000, 0.01
+    n = int(np.sqrt(n_elems / density if "sparse" in array_type.mod else n_elems))
+    arr = array_type.random((n, n), density=density, dtype=dtype)
 
     func(arr, axis=axis)  # warmup: numba compile
     benchmark(func, arr, axis=axis)