diff --git a/tests/test_inference.py b/tests/test_inference.py
index 73d05a0..d7bb02a 100644
--- a/tests/test_inference.py
+++ b/tests/test_inference.py
@@ -199,20 +199,21 @@ def test_tool_using_agent_callbacks(caplog: LogCap) -> None:
     with Client() as client:
         llm = client.llm.model(model_id)
         chat = Chat()
-        chat.add_user_message("What is the sum of 123 and 3210?")
+        # Ensure the first response is a combination of text and tool use requests
+        chat.add_user_message("First say 'Hi'. Then calculate 1 + 3 with the tool.")
         tools = [ADDITION_TOOL_SPEC]
         round_starts: list[int] = []
         round_ends: list[int] = []
         first_tokens: list[int] = []
         predictions: list[PredictionRoundResult] = []
         fragments: list[LlmPredictionFragment] = []
-        last_fragment_round_index = 0
+        fragment_round_indices: set[int] = set()
 
         def _append_fragment(f: LlmPredictionFragment, round_index: int) -> None:
-            nonlocal last_fragment_round_index
+            last_fragment_round_index = max(fragment_round_indices, default=-1)
             assert round_index >= last_fragment_round_index
             fragments.append(f)
-            last_fragment_round_index = round_index
+            fragment_round_indices.add(round_index)
 
         # TODO: Also check on_prompt_processing_progress and handling invalid messages
         # (although it isn't clear how to provoke calls to the latter without mocking)
@@ -233,8 +234,9 @@ def _append_fragment(f: LlmPredictionFragment, round_index: int) -> None:
         assert round_starts == sequential_round_indices
         assert round_ends == sequential_round_indices
         expected_token_indices = [p.round_index for p in predictions if p.content]
+        assert expected_token_indices == sequential_round_indices
         assert first_tokens == expected_token_indices
-        assert last_fragment_round_index == num_rounds - 1
+        assert fragment_round_indices == set(expected_token_indices)
         assert len(chat._messages) == 2 * num_rounds  # No tool results in last round
 
         cloned_chat = chat.copy()