diff --git a/numcodecs/compat.py b/numcodecs/compat.py
index d1844e10..355338f5 100644
--- a/numcodecs/compat.py
+++ b/numcodecs/compat.py
@@ -2,11 +2,12 @@
 import codecs
 
 import numpy as np
+import numpy.typing as npt
 
-from .ndarray_like import NDArrayLike, is_ndarray_like
+from .ndarray_like import is_ndarray_like
 
 
-def ensure_ndarray_like(buf) -> NDArrayLike:
+def ensure_ndarray_like(buf) -> npt.NDArray:
     """Convenience function to coerce `buf` to ndarray-like array.
 
     Parameters
@@ -38,7 +39,7 @@ def ensure_ndarray_like(buf) -> NDArrayLike:
             mem = memoryview(buf)
             # instantiate array from memoryview, ensures no copy
             buf = np.array(mem, copy=False)
-    return buf
+    return np.asanyarray(buf, copy=False)
 
 
 def ensure_ndarray(buf) -> np.ndarray:
@@ -63,7 +64,7 @@ def ensure_ndarray(buf) -> np.ndarray:
     return np.array(ensure_ndarray_like(buf), copy=False)
 
 
-def ensure_contiguous_ndarray_like(buf, max_buffer_size=None, flatten=True) -> NDArrayLike:
+def ensure_contiguous_ndarray_like(buf, max_buffer_size=None, flatten=True) -> npt.ArrayLike:
     """Convenience function to coerce `buf` to ndarray-like array.
     Also ensures that the returned value exports fully contiguous memory,
     and supports the new-style buffer interface. If the optional max_buffer_size is
@@ -174,7 +175,7 @@ def ensure_text(s, encoding="utf-8"):
     return s
 
 
-def ndarray_copy(src, dst) -> NDArrayLike:
+def ndarray_copy(src, dst) -> npt.NDArray:
     """Copy the contents of the array from `src` to `dst`."""
 
     if dst is None:
diff --git a/numcodecs/ndarray_like.py b/numcodecs/ndarray_like.py
index 06e15ea9..170428c5 100644
--- a/numcodecs/ndarray_like.py
+++ b/numcodecs/ndarray_like.py
@@ -1,65 +1,6 @@
-from typing import Any, ClassVar, Protocol, runtime_checkable
-
-
-class _CachedProtocolMeta(Protocol.__class__):  # type: ignore[name-defined]
-    """Custom implementation of @runtime_checkable
-
-    The native implementation of @runtime_checkable is slow,
-    see <https://github.com/zarr-developers/numcodecs/issues/379>.
-
-    This metaclass keeps an unbounded cache of the result of
-    isinstance checks using the object's class as the cache key.
-    """
-
-    _instancecheck_cache: ClassVar[dict[tuple[type, type], bool]] = {}
-
-    def __instancecheck__(self, instance):
-        key = (self, instance.__class__)
-        ret = self._instancecheck_cache.get(key)
-        if ret is None:
-            ret = super().__instancecheck__(instance)
-            self._instancecheck_cache[key] = ret
-        return ret
-
-
-@runtime_checkable
-class DType(Protocol, metaclass=_CachedProtocolMeta):
-    itemsize: int
-    name: str
-    kind: str
-
-
-@runtime_checkable
-class FlagsObj(Protocol, metaclass=_CachedProtocolMeta):
-    c_contiguous: bool
-    f_contiguous: bool
-    owndata: bool
-
-
-@runtime_checkable
-class NDArrayLike(Protocol, metaclass=_CachedProtocolMeta):
-    dtype: DType
-    shape: tuple[int, ...]
-    strides: tuple[int, ...]
-    ndim: int
-    size: int
-    itemsize: int
-    nbytes: int
-    flags: FlagsObj
-
-    def __len__(self) -> int: ...  # pragma: no cover
-
-    def __getitem__(self, key) -> Any: ...  # pragma: no cover
-
-    def __setitem__(self, key, value): ...  # pragma: no cover
-
-    def tobytes(self, order: str | None = ...) -> bytes: ...  # pragma: no cover
-
-    def reshape(self, *shape: int, order: str = ...) -> "NDArrayLike": ...  # pragma: no cover
-
-    def view(self, dtype: DType = ...) -> "NDArrayLike": ...  # pragma: no cover
+import numpy.typing as npt
 
 
 def is_ndarray_like(obj: object) -> bool:
     """Return True when `obj` is ndarray-like"""
-    return isinstance(obj, NDArrayLike)
+    return isinstance(obj, npt.ArrayLike)
diff --git a/numcodecs/tests/test_ndarray_like.py b/numcodecs/tests/test_ndarray_like.py
deleted file mode 100644
index 6c16e7db..00000000
--- a/numcodecs/tests/test_ndarray_like.py
+++ /dev/null
@@ -1,48 +0,0 @@
-import pytest
-
-from numcodecs.ndarray_like import DType, FlagsObj, NDArrayLike
-
-
-@pytest.mark.parametrize("module", ["numpy", "cupy"])
-def test_is_ndarray_like(module):
-    m = pytest.importorskip(module)
-    a = m.arange(10)
-    assert isinstance(a, NDArrayLike)
-
-
-def test_is_not_ndarray_like():
-    assert not isinstance([1, 2, 3], NDArrayLike)
-    assert not isinstance(b"1,2,3", NDArrayLike)
-
-
-@pytest.mark.parametrize("module", ["numpy", "cupy"])
-def test_is_dtype_like(module):
-    m = pytest.importorskip(module)
-    d = m.dtype("u8")
-    assert isinstance(d, DType)
-
-
-def test_is_not_dtype_like():
-    assert not isinstance([1, 2, 3], DType)
-    assert not isinstance(b"1,2,3", DType)
-
-
-@pytest.mark.parametrize("module", ["numpy", "cupy"])
-def test_is_flags_like(module):
-    m = pytest.importorskip(module)
-    d = m.arange(10).flags
-    assert isinstance(d, FlagsObj)
-
-
-def test_is_not_flags_like():
-    assert not isinstance([1, 2, 3], FlagsObj)
-    assert not isinstance(b"1,2,3", FlagsObj)
-
-
-@pytest.mark.parametrize("module", ["numpy", "cupy"])
-def test_cached_isinstance_check(module):
-    m = pytest.importorskip(module)
-    a = m.arange(10)
-    assert isinstance(a, NDArrayLike)
-    assert not isinstance(a, DType)
-    assert not isinstance(a, FlagsObj)