diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 876196b4..54daa817 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -33,7 +33,7 @@ repos:
         args:
           - --strict
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.8.1
+    rev: v0.8.6
     hooks:
       - id: ruff
       - id: ruff-format
@@ -48,7 +48,7 @@ repos:
     hooks:
       - id: toml-sort-fix
   - repo: https://github.com/rbubley/mirrors-prettier
-    rev: v3.4.1
+    rev: v3.4.2
     hooks:
       - id: prettier
   - repo: https://github.com/igorshubovych/markdownlint-cli
@@ -62,11 +62,11 @@ repos:
     hooks:
       - id: forbid-tabs
   - repo: https://github.com/rhysd/actionlint
-    rev: v1.7.4
+    rev: v1.7.6
     hooks:
       - id: actionlint
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.13.0
+    rev: v1.14.1
     hooks:
       - id: mypy
         files: ^glass/
diff --git a/glass/core/array.py b/glass/core/array.py
index 7a7dc094..4c89abdb 100644
--- a/glass/core/array.py
+++ b/glass/core/array.py
@@ -166,7 +166,7 @@ def trapezoid_product(
     y = np.interp(x, *f)
     for f_ in ff:
         y *= np.interp(x, *f_)
-    return np.trapezoid(y, x, axis=axis)  # type: ignore[no-any-return]
+    return np.atleast_1d(np.trapezoid(y, x, axis=axis))
 
 
 def cumulative_trapezoid(
diff --git a/glass/points.py b/glass/points.py
index 39f1cf23..6b9fb4c3 100644
--- a/glass/points.py
+++ b/glass/points.py
@@ -85,7 +85,7 @@ def effective_bias(
 
     """
     norm = np.trapezoid(w.wa, w.za)
-    return trapezoid_product((z, bz), (w.za, w.wa)) / norm
+    return (trapezoid_product((z, bz), (w.za, w.wa)) / norm).astype(np.float64)
 
 
 def linear_bias(
@@ -107,7 +107,7 @@ def linear_bias(
         The density contrast after biasing.
 
     """
-    return b * delta
+    return (b * delta).astype(np.float64)
 
 
 def loglinear_bias(
@@ -413,6 +413,6 @@ def position_weights(
     densities = densities / np.sum(densities, axis=0)
     # apply bias after normalisation
     if bias is not None:
-        densities = densities * bias
+        densities = (densities * bias).astype(np.float64)
     # densities now contains the relative contribution with bias applied
     return densities
diff --git a/glass/shells.py b/glass/shells.py
index a28a1295..74fe2e8a 100644
--- a/glass/shells.py
+++ b/glass/shells.py
@@ -235,8 +235,8 @@ def tophat_windows(
     ws = []
     for zmin, zmax in itertools.pairwise(zbins):
         n = max(round((zmax - zmin) / dz), 2)
-        z = np.linspace(zmin, zmax, n)
-        w = wht(z)
+        z = np.linspace(zmin, zmax, n, dtype=np.float64)
+        w = np.asarray(wht(z), dtype=np.float64)
         zeff = np.trapezoid(w * z, z) / np.trapezoid(w, z)
         ws.append(RadialWindow(z, w, zeff))
     return ws
@@ -410,7 +410,7 @@ def restrict(
     z_ = np.compress(np.greater(z, w.za[0]) & np.less(z, w.za[-1]), z)
     zr = np.union1d(w.za, z_)
     fr = ndinterp(zr, z, f, left=0.0, right=0.0) * ndinterp(zr, w.za, w.wa)
-    return zr, fr
+    return zr.astype(np.float64), fr.astype(np.float64)
 
 
 def partition(
@@ -565,7 +565,7 @@ def partition_lstsq(
     # compute the union of all given redshift grids
     zp = z
     for w in shells:
-        zp = np.union1d(zp, w.za)
+        zp = np.union1d(zp, w.za).astype(np.float64)
 
     # get extra leading axes of fz
     *dims, _ = np.shape(fz)
@@ -629,7 +629,7 @@ def partition_nnls(
     # compute the union of all given redshift grids
     zp = z
     for w in shells:
-        zp = np.union1d(zp, w.za)
+        zp = np.union1d(zp, w.za).astype(np.float64)
 
     # get extra leading axes of fz
     *dims, _ = np.shape(fz)
@@ -673,7 +673,8 @@ def partition_nnls(
     # for each dim, find non-negative weights x such that y == r @ x
     x = np.empty([len(shells), *dims])
     for i in np.ndindex(*dims):
-        x[(..., *i)] = nnls(r, y[i])
+        index = (slice(None),) + i  # noqa: RUF005
+        x[index] = nnls(r, y[i])
 
     # all done
     return x
@@ -740,9 +741,9 @@ def _uniform_grid(
 
     """
     if step is not None and num is None:
-        return np.arange(start, np.nextafter(stop + step, stop), step)
+        return np.arange(start, np.nextafter(stop + step, stop), step, dtype=np.float64)
     if step is None and num is not None:
-        return np.linspace(start, stop, num + 1)
+        return np.linspace(start, stop, num + 1, dtype=np.float64)
     msg = "exactly one of grid step size or number of steps must be given"
     raise ValueError(msg)