diff --git a/tests/realtime/test_session.py b/tests/realtime/test_session.py
index c1c919a866..9ea712a013 100644
--- a/tests/realtime/test_session.py
+++ b/tests/realtime/test_session.py
@@ -1505,6 +1505,31 @@ def __str__(self) -> str:
 
         assert _serialize_tool_output(BrokenDataclass(lock=threading.Lock())) == "broken-dataclass"
 
+    @dataclasses.dataclass
+    class ToolResult:
+        label: str
+        values: list[int]
+
+    @pytest.mark.parametrize(
+        ("value", "expected"),
+        [
+            pytest.param(None, "null", id="none"),
+            pytest.param(
+                ["hello", 1, True, None],
+                json.dumps(["hello", 1, True, None]),
+                id="list",
+            ),
+            pytest.param(
+                ToolResult(label="demo", values=[1, 2]),
+                json.dumps({"label": "demo", "values": [1, 2]}),
+                id="dataclass",
+            ),
+            pytest.param(b"abc", "b'abc'", id="bytes"),
+        ],
+    )
+    def test_serialize_tool_output_edge_cases(self, value: Any, expected: str) -> None:
+        assert _serialize_tool_output(value) == expected
+
     @pytest.mark.asyncio
     async def test_mixed_tool_types_filtering(self, mock_model, mock_agent):
         """Test that function tools and handoffs are properly separated"""