diff --git a/changes/120.feature.rst b/changes/120.feature.rst
new file mode 100644
index 00000000..c3247829
--- /dev/null
+++ b/changes/120.feature.rst
@@ -0,0 +1,2 @@
+Refactor ``AbstractDataModel`` into a Python protocol as that is how it was
+effectively used.
diff --git a/docs/source/model_library.rst b/docs/source/model_library.rst
index c065b74c..9c7f8073 100644
--- a/docs/source/model_library.rst
+++ b/docs/source/model_library.rst
@@ -4,7 +4,7 @@ Model Library
 =============
 
 `~stpipe.library.AbstractModelLibrary` is a container designed to allow efficient processing of
-collections of `~stpipe.datamodel.AbstractDataModel` instances created from an association.
+collections of `~stpipe.protocols.DataModel` instances created from an association.
 
 `~stpipe.library.AbstractModelLibrary` is an ordered collection (like a `list`) but provides:
 
@@ -234,8 +234,8 @@ allowing the `~stpipe.step.Step` to generate an :ref:`library_on_disk`
 
 ``Step.process`` can extend the above pattern to
 support additional inputs (for example a single
-`~stpipe.datamodel.AbstractDataModel` or filename containing
-a `~stpipe.datamodel.AbstractDataModel`) to allow more
+`~stpipe.protocols.DataModel` or filename containing
+a `~stpipe.protocols.DataModel`) to allow more
 flexible data processings, although some consideration
 should be given to how to handle input that does not
 contain association metadata. Does it make sense
@@ -252,9 +252,9 @@ Isolated Processing
 
 Let's say we have a `~stpipe.step.Step`, ``flux_calibration``
 that performs an operation that is only concerned with the data
-for a single `~stpipe.datamodel.AbstractDataModel` at a time.
+for a single `~stpipe.protocols.DataModel` at a time.
 This step applies a function ``calibrate_model_flux`` that
-accepts a single `~stpipe.datamodel.AbstractDataModel` and index as an input.
+accepts a single `~stpipe.protocols.DataModel` and index as an input.
 Its ``Step.process`` function can make good use of
 `~stpipe.library.AbstractModelLibrary.map_function` to apply
 this method to each model in the library.
diff --git a/src/stpipe/config_parser.py b/src/stpipe/config_parser.py
index cb1af2a1..9f376cdf 100644
--- a/src/stpipe/config_parser.py
+++ b/src/stpipe/config_parser.py
@@ -21,7 +21,7 @@
 
 from . import utilities
 from .config import StepConfig
-from .datamodel import AbstractDataModel
+from .protocols import DataModel
 from .utilities import _not_set
 
 # Configure logger
@@ -82,7 +82,7 @@ def _output_file_check(path):
 
 def _is_datamodel(value):
     """Verify that value is a DataModel."""
-    if isinstance(value, AbstractDataModel):
+    if isinstance(value, DataModel):
         return value
 
     raise VdtTypeError(value)
@@ -92,7 +92,7 @@ def _is_string_or_datamodel(value):
     """Verify that value is either a string (nominally a reference file path)
     or a DataModel (possibly one with no corresponding file.)
     """
-    if isinstance(value, AbstractDataModel):
+    if isinstance(value, DataModel):
         return value
 
     if isinstance(value, str):
diff --git a/src/stpipe/library.py b/src/stpipe/library.py
index 79f295af..d8aa4a9e 100644
--- a/src/stpipe/library.py
+++ b/src/stpipe/library.py
@@ -10,7 +10,7 @@
 
 import asdf
 
-from .datamodel import AbstractDataModel
+from .protocols import DataModel
 
 __all__ = [
     "LibraryError",
@@ -771,7 +771,7 @@ def _model_to_filename(self, model):
         return model_filename
 
     def _to_group_id(self, model_or_filename, index):
-        if isinstance(model_or_filename, AbstractDataModel):
+        if isinstance(model_or_filename, DataModel):
             getter = self._model_to_group_id
         else:
             getter = self._filename_to_group_id
diff --git a/src/stpipe/pipeline.py b/src/stpipe/pipeline.py
index 70eaf00c..4599071d 100644
--- a/src/stpipe/pipeline.py
+++ b/src/stpipe/pipeline.py
@@ -151,7 +151,7 @@ def get_config_from_reference(cls, dataset, disable=None, crds_observatory=None)
             Either a class or instance of a class derived
             from `Step`.
 
-        dataset : `stpipe.datamodel.AbstractDataModel`
+        dataset : `stpipe.protocols.DataModel`
             A model of the input file.  Metadata on this input file will
             be used by the CRDS "bestref" algorithm to obtain a reference
             file.
diff --git a/src/stpipe/datamodel.py b/src/stpipe/protocols.py
similarity index 50%
rename from src/stpipe/datamodel.py
rename to src/stpipe/protocols.py
index 008e2aaa..c58739dd 100644
--- a/src/stpipe/datamodel.py
+++ b/src/stpipe/protocols.py
@@ -1,50 +1,54 @@
-import abc
+from __future__ import annotations
 
+from abc import abstractmethod
+from typing import TYPE_CHECKING, Protocol, runtime_checkable
 
-class AbstractDataModel(abc.ABC):
+if TYPE_CHECKING:
+    from collections.abc import Callable
+    from os import PathLike
+
+
+@runtime_checkable
+class DataModel(Protocol):
     """
-    This Abstract Base Class is intended to cover multiple implementations of
-    data models so that each will be considered an appropriate subclass of this
-    class without requiring that they inherit this class.
+    This is a protocol to describe the methods and properties that define a
+    DataModel for the purposes of stpipe. This is a runtime checkable protocol
+    meaning that any object can be `isinstance` checked against this protocol
+    and will succeed even if the object does not inherit from this class.
+    Moreover, this object will act as an `abc.ABC` class if it is inherited from.
 
     Any datamodel class instance that desires to be considered an instance of
-    AbstractDataModel must implement the following methods.
+    must fully implement the protocol in order to pass the `isinstance` check.
 
     In addition, although it isn't yet checked (the best approach for supporting
     this is still being considered), such instances must have a meta.filename
     attribute.
     """
 
-    @classmethod
-    def __subclasshook__(cls, c_):
-        """
-        Pseudo subclass check based on these attributes and methods
-        """
-        if cls is AbstractDataModel:
-            mro = c_.__mro__
-            if (
-                any(hasattr(CC, "crds_observatory") for CC in mro)
-                and any(hasattr(CC, "get_crds_parameters") for CC in mro)
-                and any(hasattr(CC, "save") for CC in mro)
-            ):
-                return True
-        return False
-
     @property
-    @abc.abstractmethod
-    def crds_observatory(self):
+    @abstractmethod
+    def crds_observatory(self) -> str:
         """This should return a string identifying the observatory as CRDS expects it"""
+        ...
 
-    @abc.abstractmethod
-    def get_crds_parameters(self):
+    @property
+    @abstractmethod
+    def get_crds_parameters(self) -> dict[str, any]:
         """
         This should return a dictionary of key/value pairs corresponding to the
         parkey values CRDS is using to match reference files. Typically it returns
         all metadata simple values.
         """
+        ...
 
-    @abc.abstractmethod
-    def save(self, path, dir_path=None, *args, **kwargs):
+    @abstractmethod
+    def save(
+        self,
+        path: PathLike | Callable[..., PathLike],
+        dir_path: PathLike | None = None,
+        *args,
+        **kwargs,
+    ) -> PathLike:
         """
         Save to a file.
 
@@ -64,3 +68,4 @@ def save(self, path, dir_path=None, *args, **kwargs):
         output_path: str
             The file path the model was saved in.
         """
+        ...
diff --git a/src/stpipe/step.py b/src/stpipe/step.py
index 8776eb5f..c6e5498d 100644
--- a/src/stpipe/step.py
+++ b/src/stpipe/step.py
@@ -33,9 +33,9 @@
     DISCOURAGED_TYPES = None
 
 from . import config, config_parser, crds_client, log, utilities
-from .datamodel import AbstractDataModel
 from .format_template import FormatTemplate
 from .library import AbstractModelLibrary
+from .protocols import DataModel
 from .utilities import _not_set
 
 
@@ -398,7 +398,7 @@ def _check_args(self, args, discouraged_types, msg):
         for i, arg in enumerate(args):
             if isinstance(arg, discouraged_types):
                 self.log.error(
-                    "%s %s object.  Use an instance of AbstractDataModel instead.",
+                    "%s %s object.  Use an instance of DataModel instead.",
                     msg,
                     i,
                 )
@@ -492,7 +492,7 @@ def run(self, *args):
                                         e,
                                     )
                                 library.shelve(model, i)
-                    elif isinstance(args[0], AbstractDataModel):
+                    elif isinstance(args[0], DataModel):
                         if self.class_alias is not None:
                             if isinstance(args[0], Sequence):
                                 for model in args[0]:
@@ -506,7 +506,7 @@ def run(self, *args):
                                             "header: %s",
                                             e,
                                         )
-                            elif isinstance(args[0], AbstractDataModel):
+                            elif isinstance(args[0], DataModel):
                                 try:
                                     args[0][
                                         f"meta.cal_step.{self.class_alias}"
@@ -567,9 +567,7 @@ def run(self, *args):
                     for idx, result in enumerate(results_to_save):
                         if len(results_to_save) <= 1:
                             idx = None
-                        if isinstance(
-                            result, (AbstractDataModel | AbstractModelLibrary)
-                        ):
+                        if isinstance(result, (DataModel | AbstractModelLibrary)):
                             self.save_model(result, idx=idx)
                         elif hasattr(result, "save"):
                             try:
@@ -609,7 +607,7 @@ def finalize_result(self, result, reference_files_used):
 
         Parameters
         ----------
-        result : a datamodel that is an instance of AbstractDataModel or
+        result : a datamodel that is an instance of DataModel or
                  collections.abc.Sequence
                  One step result (potentially of many).
 
@@ -780,7 +778,7 @@ def get_ref_override(self, reference_file_type):
         """
         override_name = crds_client.get_override_name(reference_file_type)
         path = getattr(self, override_name, None)
-        if isinstance(path, AbstractDataModel):
+        if isinstance(path, DataModel):
             return path
 
         return abspath(path) if path and path != "N/A" else path
@@ -795,7 +793,7 @@ def get_reference_file(self, input_file, reference_file_type):
 
         Parameters
         ----------
-        input_file : a datamodel that is an instance of AbstractDataModel
+        input_file : a datamodel that is an instance of DataModel
             A model of the input file.  Metadata on this input file
             will be used by the CRDS "bestref" algorithm to obtain a
             reference file.
@@ -810,7 +808,7 @@ def get_reference_file(self, input_file, reference_file_type):
         """
         override = self.get_ref_override(reference_file_type)
         if override is not None:
-            if isinstance(override, AbstractDataModel):
+            if isinstance(override, DataModel):
                 self._reference_files_used.append(
                     (reference_file_type, override.override_handle)
                 )
@@ -846,7 +844,7 @@ def get_config_from_reference(cls, dataset, disable=None, crds_observatory=None)
         cls : stpipe.Step
             Either a class or instance of a class derived
             from `Step`.
-        dataset : A datamodel that is an instance of AbstractDataModel
+        dataset : A datamodel that is an instance of DataModel
             A model of the input file.  Metadata on this input file will
             be used by the CRDS "bestref" algorithm to obtain a reference
             file.
@@ -874,7 +872,7 @@ def get_config_from_reference(cls, dataset, disable=None, crds_observatory=None)
             if crds_observatory is None:
                 raise ValueError("Need a valid name for crds_observatory.")
         else:
-            # If the dataset is not an operable instance of AbstractDataModel,
+            # If the dataset is not an operable instance of DataModel,
             # log as such and return an empty config object
             try:
                 with cls._datamodels_open(dataset, asn_n_members=1) as model:
@@ -884,7 +882,7 @@ def get_config_from_reference(cls, dataset, disable=None, crds_observatory=None)
                     crds_parameters = model.get_crds_parameters()
                     crds_observatory = model.crds_observatory
             except (OSError, TypeError, ValueError):
-                logger.warning("Input dataset is not an instance of AbstractDataModel.")
+                logger.warning("Input dataset is not an instance of DataModel.")
                 disable = True
 
         # Check if retrieval should be attempted.
@@ -930,7 +928,7 @@ def set_primary_input(self, obj, exclusive=True):
 
         Parameters
         ----------
-        obj : str, pathlib.Path, or instance of AbstractDataModel
+        obj : str, pathlib.Path, or instance of DataModel
             The object to base the name on. If a datamodel,
             use Datamodel.meta.filename.
 
@@ -945,7 +943,7 @@ def set_primary_input(self, obj, exclusive=True):
         if not exclusive or parent_input_filename is None:
             if isinstance(obj, str | Path):
                 self._input_filename = str(obj)
-            elif isinstance(obj, AbstractDataModel):
+            elif isinstance(obj, DataModel):
                 try:
                     self._input_filename = obj.meta.filename
                 except AttributeError:
@@ -967,7 +965,7 @@ def save_model(
 
         Parameters
         ----------
-        model : a instance of AbstractDataModel
+        model : a instance of DataModel
             The model to save.
 
         suffix : str
@@ -1166,7 +1164,7 @@ def open_model(self, init, **kwargs):
 
         Returns
         -------
-        datamodel : instance of AbstractDataModel
+        datamodel : instance of DataModel
             Object opened as a datamodel
         """
         # Use the parent method if available, since this step
diff --git a/src/stpipe/subproc.py b/src/stpipe/subproc.py
index b9c0ad04..31c7771c 100644
--- a/src/stpipe/subproc.py
+++ b/src/stpipe/subproc.py
@@ -1,7 +1,7 @@
 import os
 import subprocess
 
-from .datamodel import AbstractDataModel
+from .protocols import DataModel
 from .step import Step
 
 
@@ -28,7 +28,7 @@ class SystemCall(Step):
     def process(self, *args):
         newargs = []
         for i, arg in enumerate(args):
-            if isinstance(arg, AbstractDataModel):
+            if isinstance(arg, DataModel):
                 filename = f"{self.qualified_name}.{i:04d}.{self.output_ext}"
                 arg.save(filename)
                 newargs.append(filename)
diff --git a/tests/test_abstract_datamodel.py b/tests/test_abstract_datamodel.py
deleted file mode 100644
index 3dcaa740..00000000
--- a/tests/test_abstract_datamodel.py
+++ /dev/null
@@ -1,57 +0,0 @@
-"""
-Test that the AbstractDataModel interface works properly
-"""
-
-import pytest
-
-from stpipe.datamodel import AbstractDataModel
-
-
-def test_roman_datamodel():
-    roman_datamodels = pytest.importorskip("roman_datamodels.datamodels")
-    from roman_datamodels.maker_utils import mk_level2_image
-
-    roman_image_tree = mk_level2_image()
-    image_model = roman_datamodels.ImageModel(roman_image_tree)
-    assert isinstance(image_model, AbstractDataModel)
-
-
-def test_jwst_datamodel():
-    jwst_datamodel = pytest.importorskip("stdatamodels.jwst.datamodels")
-    image_model = jwst_datamodel.ImageModel()
-    assert isinstance(image_model, AbstractDataModel)
-
-
-class GoodDataModel:
-    def __init__(self):
-        pass
-
-    def crds_observatory(self):
-        pass
-
-    def get_crds_parameters(self):
-        pass
-
-    def save(self):
-        pass
-
-
-class BadDataModel:
-    def __init__(self):
-        pass
-
-    def crds_observatory(self):
-        pass
-
-    def get_crds_parameters(self):
-        pass
-
-
-def test_good_datamodel():
-    gdm = GoodDataModel()
-    assert isinstance(gdm, AbstractDataModel)
-
-
-def test_bad_datamodel():
-    gdm = BadDataModel()
-    assert not isinstance(gdm, AbstractDataModel)
diff --git a/tests/test_jwst.py b/tests/test_jwst.py
new file mode 100644
index 00000000..4bb300b9
--- /dev/null
+++ b/tests/test_jwst.py
@@ -0,0 +1,31 @@
+"""
+Integration tests with JWST pipeline
+"""
+
+from inspect import getmembers, isclass
+
+import pytest
+
+from stpipe.protocols import DataModel
+
+datamodels = pytest.importorskip("stdatamodels.jwst.datamodels")
+
+
+def test_jwst_datamodel():
+    """Smoke test to ensure the JWST datamodels work with the DataModel protocol."""
+    jwst_datamodel = pytest.importorskip("stdatamodels.jwst.datamodels")
+    image_model = jwst_datamodel.ImageModel()
+    assert isinstance(image_model, DataModel)
+
+
+@pytest.mark.parametrize(
+    "model",
+    [
+        model[1]
+        for model in getmembers(datamodels, isclass)
+        if issubclass(model[1], datamodels.JwstDataModel)
+    ],
+)
+def test_datamodel(model):
+    """Test that all JWST datamodels work with the DataModel protocol."""
+    assert isinstance(model(), DataModel)
diff --git a/tests/test_library.py b/tests/test_library.py
index 5a7e0853..3d142e1c 100644
--- a/tests/test_library.py
+++ b/tests/test_library.py
@@ -8,7 +8,6 @@
 import asdf
 import pytest
 
-from stpipe.datamodel import AbstractDataModel
 from stpipe.library import (
     AbstractModelLibrary,
     BorrowError,
@@ -16,6 +15,7 @@
     NoGroupID,
     _Ledger,
 )
+from stpipe.protocols import DataModel
 
 _GROUP_IDS = ["1", "1", "2"]
 _N_MODELS = len(_GROUP_IDS)
@@ -38,7 +38,7 @@ class _Meta:
     pass
 
 
-class DataModel:
+class ExampleDataModel:
     def __init__(self, **kwargs):
         self.meta = _Meta()
         self.meta.__dict__.update(kwargs)
@@ -57,7 +57,7 @@ def save(self, path, **kwargs):
 
 def _load_model(filename):
     with asdf.open(filename) as af:
-        return DataModel(**af.tree)
+        return ExampleDataModel(**af.tree)
 
 
 class ModelLibrary(AbstractModelLibrary):
@@ -113,7 +113,7 @@ def example_models():
     """
     models = []
     for i in range(_N_MODELS):
-        m = DataModel(group_id=_GROUP_IDS[i], index=i)
+        m = ExampleDataModel(group_id=_GROUP_IDS[i], index=i)
         m.meta.filename = f"{i}.asdf"
         models.append(m)
     return models
@@ -219,7 +219,7 @@ def test_init_from_models_no_ondisk(example_models):
         ModelLibrary(example_models, on_disk=True)
 
 
-@pytest.mark.parametrize("invalid", (None, ModelLibrary([]), DataModel()))
+@pytest.mark.parametrize("invalid", (None, ModelLibrary([]), ExampleDataModel()))
 def test_invalid_init(invalid):
     """
     Test that some unsupported init values produce errors.
@@ -408,7 +408,7 @@ def test_closed_library_model_shelve(example_library):
     an error.
     """
     with pytest.raises(ClosedLibraryError, match="ModelLibrary is not open"):
-        example_library.shelve(DataModel(), 0)
+        example_library.shelve(ExampleDataModel(), 0)
 
 
 def test_closed_library_model_iter(example_library):
@@ -577,7 +577,7 @@ def test_shelve_unknown_model(example_library, use_index):
     with lib_ctx:  # to catch the error for the un-returned model
         with example_library:
             example_library.borrow(0)
-            new_model = DataModel()
+            new_model = ExampleDataModel()
 
             if use_index:
                 ctx = contextlib.nullcontext()
@@ -758,7 +758,7 @@ def test_ledger():
     based on index and index based on models.
     """
     ledger = _Ledger()
-    model = DataModel()
+    model = ExampleDataModel()
     ledger[0] = model
     assert ledger[0] == model
     assert ledger[model] == 0
@@ -773,10 +773,21 @@ def test_ledger():
 def test_library_datamodel_relationship():
     """
     Smoke test to make sure the relationship between
-    AbstractModelLibrary and AbstractDataModel doesn't
+    AbstractModelLibrary and DataModel doesn't
     change.
     """
-    assert not issubclass(AbstractModelLibrary, AbstractDataModel)
+    dm = ExampleDataModel()
+    lib = ModelLibrary([])
+
+    assert isinstance(dm, DataModel)
+    assert isinstance(lib, AbstractModelLibrary)
+
+    assert not isinstance(dm, AbstractModelLibrary)
+    assert not isinstance(lib, DataModel)
+
+    # Protocols don't support issubclass for complex reasons
+    # so we check the relationship with isinstance on instances instead
+    # assert not issubclass(AbstractModelLibrary, DataModel)
 
 
 def test_library_is_not_sequence():
diff --git a/tests/test_protocols.py b/tests/test_protocols.py
new file mode 100644
index 00000000..eb4239a7
--- /dev/null
+++ b/tests/test_protocols.py
@@ -0,0 +1,40 @@
+"""
+Test that the DataModel interface of the protocol works properly
+"""
+
+from stpipe.protocols import DataModel
+
+
+class GoodDataModel:
+    def __init__(self):
+        pass
+
+    def crds_observatory(self):
+        pass
+
+    def get_crds_parameters(self):
+        pass
+
+    def save(self):
+        pass
+
+
+class BadDataModel:
+    def __init__(self):
+        pass
+
+    def crds_observatory(self):
+        pass
+
+    def get_crds_parameters(self):
+        pass
+
+
+def test_good_datamodel():
+    gdm = GoodDataModel()
+    assert isinstance(gdm, DataModel)
+
+
+def test_bad_datamodel():
+    gdm = BadDataModel()
+    assert not isinstance(gdm, DataModel)
diff --git a/tests/test_roman.py b/tests/test_roman.py
new file mode 100644
index 00000000..7e7c31ff
--- /dev/null
+++ b/tests/test_roman.py
@@ -0,0 +1,35 @@
+"""
+Integration tests with Roman pipeline
+"""
+
+from inspect import getmembers, isclass
+
+import pytest
+
+from stpipe.protocols import DataModel
+
+datamodels = pytest.importorskip("roman_datamodels.datamodels")
+
+
+def test_roman_datamodel():
+    """Smoke test to ensure the Roman datamodels work with the DataModel protocol."""
+    roman_datamodels = pytest.importorskip("roman_datamodels.datamodels")
+    from roman_datamodels.maker_utils import mk_level2_image
+
+    roman_image_tree = mk_level2_image()
+    image_model = roman_datamodels.ImageModel(roman_image_tree)
+    assert isinstance(image_model, DataModel)
+
+
+@pytest.mark.parametrize(
+    "model",
+    [
+        model[1]
+        for model in getmembers(datamodels, isclass)
+        if model[1] != datamodels.DataModel
+        and issubclass(model[1], datamodels.DataModel)
+    ],
+)
+def test_datamodel(model):
+    """Test that all Roman datamodels work with the DataModel protocol."""
+    assert isinstance(model(), DataModel)