diff --git a/pandas-stubs/core/frame.pyi b/pandas-stubs/core/frame.pyi
index 6cd3236bb..f413e6110 100644
--- a/pandas-stubs/core/frame.pyi
+++ b/pandas-stubs/core/frame.pyi
@@ -1009,7 +1009,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: Scalar,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1022,7 +1022,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: DatetimeIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1035,7 +1035,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: TimedeltaIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1048,7 +1048,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: PeriodIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1061,7 +1061,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: IntervalIndex[IntervalT],
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1074,7 +1074,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: MultiIndex | GroupByObjectNonScalar | None = ...,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1087,7 +1087,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: Series[SeriesByT],
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -1100,7 +1100,7 @@ class DataFrame(NDFrame, OpsMixin):
         self,
         by: CategoricalIndex | Index | Series,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
diff --git a/pandas-stubs/core/series.pyi b/pandas-stubs/core/series.pyi
index aeee0fecd..2a42e8a3d 100644
--- a/pandas-stubs/core/series.pyi
+++ b/pandas-stubs/core/series.pyi
@@ -117,6 +117,7 @@ from pandas._typing import (
     HashableT3,
     IgnoreRaise,
     IndexingInt,
+    IndexLabel,
     IntDtypeArg,
     InterpolateOptions,
     IntervalClosedType,
@@ -547,7 +548,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: Scalar,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -560,7 +561,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: DatetimeIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -573,7 +574,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: TimedeltaIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -586,7 +587,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: PeriodIndex,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -599,7 +600,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: IntervalIndex[IntervalT],
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -612,7 +613,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: MultiIndex | GroupByObjectNonScalar = ...,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -625,7 +626,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: Series[SeriesByT],
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
@@ -638,7 +639,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self,
         by: CategoricalIndex | Index | Series,
         axis: AxisIndex = ...,
-        level: Level | None = ...,
+        level: IndexLabel | None = ...,
         as_index: _bool = ...,
         sort: _bool = ...,
         group_keys: _bool = ...,
diff --git a/tests/test_frame.py b/tests/test_frame.py
index a25e9b683..bdd9bf491 100644
--- a/tests/test_frame.py
+++ b/tests/test_frame.py
@@ -1073,6 +1073,21 @@ def test_types_groupby_iter() -> None:
     )
 
 
+def test_types_groupby_level() -> None:
+    # GH 836
+    data = {
+        "col1": [0, 0, 0],
+        "col2": [0, 1, 0],
+        "col3": [1, 2, 3],
+        "col4": [1, 2, 3],
+    }
+    df = pd.DataFrame(data=data).set_index(["col1", "col2", "col3"])
+    check(
+        assert_type(df.groupby(level=["col1", "col2"]).sum(), pd.DataFrame),
+        pd.DataFrame,
+    )
+
+
 def test_types_merge() -> None:
     df = pd.DataFrame(data={"col1": [1, 1, 2], "col2": [3, 4, 5]})
     df2 = pd.DataFrame(data={"col1": [1, 1, 2], "col2": [0, 1, 0]})
diff --git a/tests/test_series.py b/tests/test_series.py
index 72ecd0cbe..ba6ae57f7 100644
--- a/tests/test_series.py
+++ b/tests/test_series.py
@@ -439,6 +439,19 @@ def test_types_max() -> None:
     s.max(skipna=False)
 
 
+def test_types_groupby_level() -> None:
+    # GH 836
+    index = pd.MultiIndex.from_tuples(
+        [(0, 0, 1), (0, 1, 2), (0, 0, 3)], names=["col1", "col2", "col3"]
+    )
+    s = pd.Series([1, 2, 3], index=index)
+    check(
+        assert_type(s.groupby(level=["col1", "col2"]).sum(), "pd.Series[int]"),
+        pd.Series,
+        np.integer,
+    )
+
+
 def test_types_quantile() -> None:
     s = pd.Series([1, 2, 3, 10])
     s.quantile([0.25, 0.5])