diff --git a/src/ntops/kernels/cos.py b/src/ntops/kernels/cos.py
new file mode 100644
index 0000000..72fcbcf
--- /dev/null
+++ b/src/ntops/kernels/cos.py
@@ -0,0 +1,16 @@
+import functools
+
+import ninetoothed
+import ninetoothed.language as ntl
+from ninetoothed import Tensor
+
+from ntops.kernels.element_wise import arrangement
+
+
+def application(input, output):
+    output = ntl.cos(input)  # noqa: F841
+
+
+@functools.cache
+def make(ndim):
+    return ninetoothed.make(arrangement, application, (Tensor(ndim), Tensor(ndim)))
diff --git a/src/ntops/torch.py b/src/ntops/torch.py
index 6d4e487..c27c2af 100644
--- a/src/ntops/torch.py
+++ b/src/ntops/torch.py
@@ -4,6 +4,7 @@
 import ntops.kernels.add
 import ntops.kernels.addmm
 import ntops.kernels.bmm
+import ntops.kernels.cos
 import ntops.kernels.div
 import ntops.kernels.exp
 import ntops.kernels.gelu
@@ -62,6 +63,17 @@ def bmm(input, mat2, *, out=None):
     return out
 
 
+def cos(input, *, out=None):
+    if out is None:
+        out = torch.empty_like(input)
+
+    kernel = ntops.kernels.cos.make(input.ndim)
+
+    kernel(input, out)
+
+    return out
+
+
 def div(input, other, *, rounding_mode=None, out=None):
     if out is None:
         out = torch.empty_like(input)
diff --git a/tests/test_cos.py b/tests/test_cos.py
new file mode 100644
index 0000000..6246ac1
--- /dev/null
+++ b/tests/test_cos.py
@@ -0,0 +1,23 @@
+import pytest
+import torch
+
+import ntops.torch
+from tests.skippers import skip_if_cuda_not_available
+from tests.utils import generate_arguments
+
+
+@skip_if_cuda_not_available
+@pytest.mark.parametrize(*generate_arguments())
+def test_cuda(shape, dtype, atol, rtol):
+    # TODO: Test for `float16` later.
+    if dtype is torch.float16:
+        return
+
+    device = "cuda"
+
+    input = torch.randn(shape, dtype=dtype, device=device)
+
+    ninetoothed_output = ntops.torch.cos(input)
+    reference_output = torch.cos(input)
+
+    assert torch.allclose(ninetoothed_output, reference_output, atol=atol, rtol=rtol)