diff --git a/litellm/llms/dashscope/cost_calculator.py b/litellm/llms/dashscope/cost_calculator.py
index 8bb7f605b82..2bfaad21259 100644
--- a/litellm/llms/dashscope/cost_calculator.py
+++ b/litellm/llms/dashscope/cost_calculator.py
@@ -46,6 +46,23 @@ def _extract_token_breakdown(usage: Usage) -> TokenBreakdown:
     )
 
 
+def _resolve_tier_cost_per_token(
+    tier: dict, cost_key: str, fallback_cost_key: Optional[str]
+) -> float:
+    """Resolve a tier's per-token cost.
+
+    An explicit 0.0 is a real price (e.g. a free-cache-read tier) and must not
+    be treated as "missing". Only fall back to ``fallback_cost_key`` when the
+    primary key is absent. This mirrors the flat-pricing path in
+    ``_calculate_prompt_cost`` / ``_calculate_completion_cost``, which already
+    guards with ``is None``.
+    """
+    cost = tier.get(cost_key)
+    if cost is None and fallback_cost_key is not None:
+        cost = tier.get(fallback_cost_key)
+    return float(cost) if cost is not None else 0.0
+
+
 def _calculate_tiered_cost(
     tokens: int,
     tiered_pricing: List[dict],
@@ -105,7 +122,9 @@ def _calculate_tiered_cost(
 
         if tier_end > tier_start:
             tokens_in_tier = tier_end - tier_start
-            cost_per_token = tier.get(cost_key) or tier.get(fallback_cost_key, 0)
+            cost_per_token = _resolve_tier_cost_per_token(
+                tier, cost_key, fallback_cost_key
+            )
             total_cost += tokens_in_tier * cost_per_token
             tokens_processed = tier_end
 
@@ -114,7 +133,9 @@ def _calculate_tiered_cost(
     if tokens_processed < tokens and sorted_tiers:
         last_tier = sorted_tiers[-1]
         remaining_tokens = tokens - tokens_processed
-        cost_per_token = last_tier.get(cost_key) or last_tier.get(fallback_cost_key, 0)
+        cost_per_token = _resolve_tier_cost_per_token(
+            last_tier, cost_key, fallback_cost_key
+        )
         total_cost += remaining_tokens * cost_per_token
 
     return total_cost
diff --git a/tests/test_litellm/llms/dashscope/test_dashscope_cost_calculator.py b/tests/test_litellm/llms/dashscope/test_dashscope_cost_calculator.py
index 05279202e8e..0554a68f107 100644
--- a/tests/test_litellm/llms/dashscope/test_dashscope_cost_calculator.py
+++ b/tests/test_litellm/llms/dashscope/test_dashscope_cost_calculator.py
@@ -8,7 +8,6 @@
 - Correctly calculates costs for token counts exceeding the highest defined tier.
 """
 
-import json
 import math
 import os
 import sys
@@ -20,6 +19,7 @@
 
 import litellm
 from litellm.llms.dashscope.cost_calculator import (
+    _calculate_tiered_cost,
     cost_per_token as dashscope_cost_per_token,
 )
 from litellm.types.utils import Usage, PromptTokensDetailsWrapper
@@ -157,3 +157,53 @@ def test_dashscope_tiered_pricing_exceeding_highest_tier(self):
         )
 
         assert math.isclose(prompt_cost, expected_prompt_cost, rel_tol=1e-10)
+
+    def test_dashscope_tiered_zero_cost_not_overridden_by_fallback(self):
+        """
+        A tier may legitimately price cached reads (or reasoning tokens) at 0.0,
+        e.g. a free-cache-read tier. The tiered calculator must treat an explicit
+        0.0 as a real price, not as "missing" and fall back to the full input
+        rate. This mirrors the flat-pricing path, which already guards on `None`.
+        """
+        tiered_pricing = [
+            {
+                "range": [0, 256000],
+                "input_cost_per_token": 2e-7,
+                "cache_read_input_token_cost": 0.0,  # free cache reads in this tier
+            }
+        ]
+
+        cost = _calculate_tiered_cost(
+            tokens=10000,
+            tiered_pricing=tiered_pricing,
+            cost_key="cache_read_input_token_cost",
+            fallback_cost_key="input_cost_per_token",
+        )
+
+        # 10,000 cached tokens priced at $0.0/token must cost $0.0, not the
+        # input rate (which would be 10000 * 2e-7 = $0.002).
+        assert cost == 0.0
+
+    def test_dashscope_tiered_zero_cost_applies_to_overflow_tokens(self):
+        """
+        The same 0.0-is-real-price rule must hold on the overflow path that
+        charges tokens exceeding the highest defined tier at the last tier's
+        rate.
+        """
+        tiered_pricing = [
+            {
+                "range": [0, 100],
+                "input_cost_per_token": 2e-7,
+                "cache_read_input_token_cost": 0.0,
+            }
+        ]
+
+        # 150 cached tokens: 100 in-tier + 50 overflow, both at $0.0.
+        cost = _calculate_tiered_cost(
+            tokens=150,
+            tiered_pricing=tiered_pricing,
+            cost_key="cache_read_input_token_cost",
+            fallback_cost_key="input_cost_per_token",
+        )
+
+        assert cost == 0.0