diff --git a/tests/python_package_test/test_engine.py b/tests/python_package_test/test_engine.py
index 7d7ae5c9fb01..7469c376e070 100644
--- a/tests/python_package_test/test_engine.py
+++ b/tests/python_package_test/test_engine.py
@@ -70,6 +70,11 @@ def categorize(continuous_x):
     return np.digitize(continuous_x, bins=np.arange(0, 1, 0.01))
 
 
+@pytest.fixture
+def model_path_txt(tmp_path):
+    return str(tmp_path / "lgb.model")
+
+
 def test_binary():
     X, y = load_breast_cancer(return_X_y=True)
     X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)
@@ -1258,14 +1263,13 @@ def test_cv():
     np.testing.assert_allclose(cv_res_lambda["valid ndcg@3-mean"], cv_res_lambda_obj["valid ndcg@3-mean"])
 
 
-def test_cv_works_with_init_model(tmp_path):
+def test_cv_works_with_init_model(model_path_txt):
     X, y = make_synthetic_regression()
     params = {"objective": "regression", "verbose": -1}
     num_train_rounds = 2
     lgb_train = lgb.Dataset(X, y, free_raw_data=False)
     bst = lgb.train(params=params, train_set=lgb_train, num_boost_round=num_train_rounds)
     preds_raw = bst.predict(X, raw_score=True)
-    model_path_txt = str(tmp_path / "lgb.model")
     bst.save_model(model_path_txt)
 
     num_cv_rounds = 5
@@ -1349,7 +1353,7 @@ def test_cvbooster():
     assert ret < 0.15
 
 
-def test_cvbooster_save_load(tmp_path):
+def test_cvbooster_save_load(model_path_txt):
     X, y = load_breast_cancer(return_X_y=True)
     X_train, X_test, y_train, _ = train_test_split(X, y, test_size=0.1, random_state=42)
     params = {
@@ -1372,8 +1376,6 @@ def test_cvbooster_save_load(tmp_path):
     preds = cvbooster.predict(X_test)
     best_iteration = cvbooster.best_iteration
 
-    model_path_txt = str(tmp_path / "lgb.model")
-
     cvbooster.save_model(model_path_txt)
     model_string = cvbooster.model_to_string()
     del cvbooster
@@ -1432,7 +1434,7 @@ def test_feature_name():
     assert feature_names == gbm.feature_name()
 
 
-def test_feature_name_with_non_ascii(rng, tmp_path):
+def test_feature_name_with_non_ascii(rng, model_path_txt):
     X_train = rng.normal(size=(100, 4))
     y_train = rng.normal(size=(100,))
     # This has non-ascii strings.
@@ -1442,7 +1444,6 @@ def test_feature_name_with_non_ascii(rng, tmp_path):
 
     gbm = lgb.train(params, lgb_train, num_boost_round=5)
     assert feature_names == gbm.feature_name()
-    model_path_txt = str(tmp_path / "lgb.model")
     gbm.save_model(model_path_txt)
 
     gbm2 = lgb.Booster(model_file=model_path_txt)
@@ -1498,7 +1499,7 @@ def test_parameters_are_loaded_from_model_file(tmp_path, capsys, rng):
     np.testing.assert_allclose(preds, orig_preds)
 
 
-def test_save_load_copy_pickle(tmp_path):
+def test_save_load_copy_pickle(model_path_txt, tmp_path):
     def train_and_predict(init_model=None, return_model=False):
         X, y = make_synthetic_regression()
         X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)
@@ -1510,7 +1511,6 @@ def train_and_predict(init_model=None, return_model=False):
     gbm = train_and_predict(return_model=True)
     ret_origin = train_and_predict(init_model=gbm)
     other_ret = []
-    model_path_txt = str(tmp_path / "lgb.model")
     gbm.save_model(model_path_txt)
     with open(model_path_txt) as f:  # check all params are logged into model file correctly
         assert f.read().find("[num_iterations: 10]") != -1