diff --git a/docs/changes/newsfragments/8552.underthehood b/docs/changes/newsfragments/8552.underthehood
new file mode 100644
index 00000000000..791c2c53b64
--- /dev/null
+++ b/docs/changes/newsfragments/8552.underthehood
@@ -0,0 +1,2 @@
+The test suite no longer depends on ``deepdiff``. The tests that used it now compare the data directly
+using ``InterDependencies`` equality and ``numpy.testing.assert_array_equal``.
diff --git a/pyproject.toml b/pyproject.toml
index 149388ced49..69005f47889 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -65,7 +65,6 @@ minicircuits_rudat = ["pywinusb>=0.4.2"]
 loop = ["qcodes_loop>=0.1.2"]
 test = [
     "coverage[toml]>=7.10.5",
-    "deepdiff>=8.6.1",
     "hypothesis>=6.130.0",
     "lxml>=4.9.4",
     "lxml-stubs>=0.4.0",
diff --git a/requirements.txt b/requirements.txt
index 6ee0dd40af7..10e5d9f0c94 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -27,8 +27,6 @@ broadbean==0.15.0
     # via
     #   qcodes (pyproject.toml)
     #   qcodes
-cachebox==5.2.3
-    # via deepdiff
 certifi==2026.7.22
     # via requests
 cf-xarray==0.11.3
@@ -70,8 +68,6 @@ debugpy==1.8.22
     # via ipykernel
 decorator==5.3.1
     # via ipython
-deepdiff==9.1.0
-    # via qcodes (pyproject.toml)
 defusedxml==0.7.1
     # via nbconvert
 docutils==0.21.2
@@ -241,8 +237,6 @@ opentelemetry-api==1.44.0
     # via
     #   qcodes (pyproject.toml)
     #   qcodes
-orderly-set==5.5.0
-    # via deepdiff
 packaging==26.3
     # via
     #   qcodes (pyproject.toml)
diff --git a/tests/dataset/test_converters.py b/tests/dataset/test_converters.py
index a832f736963..579fda90805 100644
--- a/tests/dataset/test_converters.py
+++ b/tests/dataset/test_converters.py
@@ -1,5 +1,3 @@
-from deepdiff import DeepDiff  # type: ignore[import-untyped]
-
 from qcodes.dataset.descriptions.rundescriber import RunDescriber
 from qcodes.dataset.descriptions.versioning.converters import (
     new_to_old,
@@ -66,11 +64,11 @@ def _assert_dicts_are_related_as_expected(v0, v1, v2):
     assert v1["version"] == 1
     assert len(v1) == 2
 
-    # conversion does not preserve order in the dict so use deepdiff to compare
-    assert (
-        DeepDiff(v2["interdependencies"], v0["interdependencies"], ignore_order=True)
-        == {}
-    )
+    # conversion does not preserve the order of the paramspecs so compare
+    # via InterDependencies whose equality ignores that order
+    assert InterDependencies._from_dict(
+        v2["interdependencies"]
+    ) == InterDependencies._from_dict(v0["interdependencies"])
     assert v2["interdependencies_"] == v1["interdependencies"]
     assert v2["version"] == 2
     assert len(v2) == 3
@@ -93,8 +91,24 @@ def test_construct_current_rundescriber_from_v0(some_paramspecs) -> None:
         version=3,
         shapes=None,
     )
-    assert DeepDiff(rds1._to_dict(), expected_v3_dict, ignore_order=True) == {}
-    assert DeepDiff(rds_upgraded._to_dict(), expected_v3_dict, ignore_order=True) == {}
+    _assert_v3_dicts_equal_up_to_paramspec_order(rds1._to_dict(), expected_v3_dict)
+    _assert_v3_dicts_equal_up_to_paramspec_order(
+        rds_upgraded._to_dict(), expected_v3_dict
+    )
+
+
+def _assert_v3_dicts_equal_up_to_paramspec_order(
+    actual: RunDescriberV3Dict, expected: RunDescriberV3Dict
+) -> None:
+    # conversion from v0 does not preserve the order of the old style
+    # paramspecs so compare those via InterDependencies whose equality
+    # ignores that order. All other entries must match exactly.
+    assert InterDependencies._from_dict(
+        actual["interdependencies"]
+    ) == InterDependencies._from_dict(expected["interdependencies"])
+    assert {k: v for k, v in actual.items() if k != "interdependencies"} == {
+        k: v for k, v in expected.items() if k != "interdependencies"
+    }
 
 
 def test_construct_current_rundescriber_from_v1(some_interdeps) -> None:
diff --git a/tests/dataset/test_dataset_in_memory.py b/tests/dataset/test_dataset_in_memory.py
index 8a737a5c109..df00b01c0cb 100644
--- a/tests/dataset/test_dataset_in_memory.py
+++ b/tests/dataset/test_dataset_in_memory.py
@@ -9,9 +9,8 @@
 import numpy as np
 import pytest
 import xarray as xr
-from deepdiff import DeepDiff  # type: ignore[import-untyped]
 from hypothesis import HealthCheck, given, settings
-from numpy.testing import assert_almost_equal
+from numpy.testing import assert_almost_equal, assert_array_equal
 
 import qcodes
 from qcodes.dataset import Measurement, load_by_id, load_by_run_spec
@@ -727,6 +726,10 @@ def test_dataset_in_mem_with_inferred_parameters(
     cache_data = ds.cache.data()
 
     assert set(param_data.keys()) == set(cache_data.keys())
-    assert set(param_data["dependent"].keys()) == set(cache_data["dependent"].keys())
-
-    assert DeepDiff(param_data, cache_data, ignore_nan_inequality=True) == {}
+    for outer_key, param_subdict in param_data.items():
+        cache_subdict = cache_data[outer_key]
+        assert set(param_subdict.keys()) == set(cache_subdict.keys())
+        for inner_key, param_array in param_subdict.items():
+            # assert_array_equal treats NaNs in the same positions as equal
+            # and strict=True also requires matching shape and dtype
+            assert_array_equal(param_array, cache_subdict[inner_key], strict=True)