diff --git a/tests/integ/test_tools.py b/tests/integ/test_tools.py
index 42c8e62e..09b8ba5c 100644
--- a/tests/integ/test_tools.py
+++ b/tests/integ/test_tools.py
@@ -133,7 +133,7 @@ def test_florence2_phrase_grounding_fine_tune_id():
         fine_tune_id=FINE_TUNE_ID,
     )
     # this calls a fine-tuned florence2 model which is going to be worse at this task
-    assert 14 <= len(result) <= 26
+    assert 13 <= len(result) <= 26
     assert [res["label"] for res in result] == ["coin"] * len(result)
     assert all([all([0 <= x <= 1 for x in obj["bbox"]]) for obj in result])