diff --git a/litellm/router.py b/litellm/router.py
index 53301979b2f..4b58629e38b 100644
--- a/litellm/router.py
+++ b/litellm/router.py
@@ -475,6 +475,15 @@ _SESSION_ADAPTER: Final = TypeAdapter(Mapping[str, object])
 _SILENT_MODEL_ADAPTER: Final = TypeAdapter(str | list[str])
 
 
+def _untried_fallback_target_exists(chain: Sequence[object] | None, kwargs: Mapping[str, Any]) -> bool:
+    """
+    Whether a resolved fallback chain still names a target this request can try.
+    `has_unattempted_fallback_target` reports an unattempted target for an empty chain
+    whenever the request carries no attempt record yet, so guard the empty case here.
+    """
+    return bool(chain) and has_unattempted_fallback_target(chain, kwargs)
+
+
 def _as_retry_skipped_deployment_ids(value: object) -> tuple[str, ...]:
     return tuple(item for item in value if isinstance(item, str)) if isinstance(value, tuple) else ()
 
@@ -8514,7 +8523,7 @@ class Router:
             fallbacks=fallbacks,
             lookup_groups=fallback_lookup_groups(kwargs, model_group),
         )
-        return has_unattempted_fallback_target(resolved, kwargs)
+        return _untried_fallback_target_exists(resolved, kwargs)
 
     def _anthropic_messages_order_levels(self, model_group: str, kwargs: Mapping[str, Any]) -> tuple[int, ...]:
         """
@@ -8583,7 +8592,7 @@ class Router:
         if model_group is None or fallbacks_disabled_for_request(kwargs):
             return False
         if _check_non_standard_fallback_format(fallbacks=fallbacks):
-            return has_unattempted_fallback_target(fallbacks, kwargs)
+            return _untried_fallback_target_exists(fallbacks, kwargs)
         dedicated_fallbacks: Final = (
             context_window_fallbacks
             if isinstance(error, litellm.ContextWindowExceededError)
@@ -8592,7 +8601,7 @@ class Router:
             else None
         )
         if dedicated_fallbacks is not None:
-            return has_unattempted_fallback_target(
+            return _untried_fallback_target_exists(
                 self._get_fallback_model_group_for_lookup_groups(
                     fallbacks=dedicated_fallbacks,
                     lookup_groups=fallback_lookup_groups(kwargs, model_group),
diff --git a/tests/unit/test_router_retry_backoff_headers.py b/tests/unit/test_router_retry_backoff_headers.py
index 75413ca633c..487fac08172 100644
--- a/tests/unit/test_router_retry_backoff_headers.py
+++ b/tests/unit/test_router_retry_backoff_headers.py
@@ -12,6 +12,7 @@ import pytest
 import litellm
 from litellm import Router
 from litellm.constants import MAX_RETRY_DELAY
+from litellm.router import _untried_fallback_target_exists
 from litellm.router_utils.fallback_event_handlers import AttemptedFallbackTargets, record_disable_fallbacks
 from litellm.types.router import RetryPolicy
 
@@ -78,6 +79,25 @@ async def test_fallback_configured_for_another_group_keeps_retry_backoff():
         )
 
 
+@pytest.mark.asyncio
+async def test_empty_fallback_chain_keeps_retry_backoff():
+    """A chain configured as an empty list names no target, so the request has nothing to
+    fall back to and the retries must still space themselves out."""
+    router: Final = _router_with_single_failing_deployment(fallbacks=[{"primary": []}])
+
+    with pytest.raises(asyncio.TimeoutError):
+        await asyncio.wait_for(
+            router.acompletion(model="primary", messages=[{"role": "user", "content": "Hello"}]),
+            timeout=_BACKOFF_DETECTION_TIMEOUT,
+        )
+
+
+def test_untried_fallback_target_exists_is_false_for_a_chain_with_no_entries():
+    assert _untried_fallback_target_exists(["backup"], {}) is True
+    assert _untried_fallback_target_exists([], {}) is False
+    assert _untried_fallback_target_exists(None, {}) is False
+
+
 @pytest.mark.asyncio
 async def test_client_side_fallback_list_does_not_back_off_before_falling_back():
     router: Final = _router_with_single_failing_deployment(fallbacks=[])