diff --git a/mtenn/tests/test_combination.py b/mtenn/tests/test_combination.py
index 4798588..13355c3 100644
--- a/mtenn/tests/test_combination.py
+++ b/mtenn/tests/test_combination.py
@@ -11,9 +11,7 @@
 @pytest.fixture()
 def models_and_inputs():
     model_test = SchNet(
-        PygSchNet(
-            hidden_channels=16, num_filters=16, num_interactions=2, num_gaussians=2
-        )
+        PygSchNet(hidden_channels=2, num_filters=2, num_interactions=2, num_gaussians=2)
     )
     model_ref = deepcopy(model_test)
     model_ref = SchNet.get_model(model_ref, strategy="complex")
@@ -56,7 +54,7 @@ def test_mean_combination(models_and_inputs):
     ref_param_dict = dict(model_ref.named_parameters())
     assert all(
         [
-            np.allclose(p.grad, ref_param_dict[n].grad, atol=1e-7)
+            np.allclose(p.grad, ref_param_dict[n].grad, atol=5e-7)
             for n, p in model_test.named_parameters()
         ]
     )
@@ -88,7 +86,7 @@ def test_max_combination(models_and_inputs):
     ref_param_dict = dict(model_ref.named_parameters())
     assert all(
         [
-            np.allclose(p.grad, ref_param_dict[n].grad, atol=1e-7)
+            np.allclose(p.grad, ref_param_dict[n].grad, atol=5e-7)
             for n, p in model_test.named_parameters()
         ]
     )
@@ -118,7 +116,7 @@ def test_boltzmann_combination(models_and_inputs):
     ref_param_dict = dict(model_ref.named_parameters())
     assert all(
         [
-            np.allclose(p.grad, ref_param_dict[n].grad, atol=1e-7)
+            np.allclose(p.grad, ref_param_dict[n].grad, atol=5e-7)
             for n, p in model_test.named_parameters()
         ]
     )