diff --git a/litellm/llms/bedrock/batches/transformation.py b/litellm/llms/bedrock/batches/transformation.py
index 4f74e3f7035..ae0f8c5935b 100644
--- a/litellm/llms/bedrock/batches/transformation.py
+++ b/litellm/llms/bedrock/batches/transformation.py
@@ -63,7 +63,7 @@ def _validate_bedrock_tags(raw_tags: object) -> list[BedrockTag]:
 
 def titan_embedding_usage_from_batch_output(model_output: Mapping[str, object]) -> Usage | None:
     """Titan embedding batch lines report usage as a top-level inputTextTokenCount, not a usage block."""
-    if "embedding" not in model_output:
+    if "embedding" not in model_output and "embeddingsByType" not in model_output:
         return None
     input_text_token_count: Final = model_output.get("inputTextTokenCount")
     if isinstance(input_text_token_count, bool) or not isinstance(input_text_token_count, int):
diff --git a/tests/test_litellm/batches/test_batch_utils.py b/tests/test_litellm/batches/test_batch_utils.py
index a2811864519..9a089112c70 100644
--- a/tests/test_litellm/batches/test_batch_utils.py
+++ b/tests/test_litellm/batches/test_batch_utils.py
@@ -1763,9 +1763,10 @@ def test_bedrock_titan_embedding_batch_usage_is_parsed():
 
 
 def test_bedrock_titan_embedding_batch_is_billed():
+    """Binary embedding rows carry only embeddingsByType and must bill like float rows."""
     rows = [
-        {"recordId": str(i), "modelOutput": {"embedding": [0.1], "inputTextTokenCount": count}}
-        for i, count in enumerate((10, 7))
+        {"recordId": "0", "modelOutput": {"embedding": [0.1], "inputTextTokenCount": 10}},
+        {"recordId": "1", "modelOutput": {"embeddingsByType": {"binary": [1, 0]}, "inputTextTokenCount": 7}},
     ]
     result = bu._aggregate_batch_cost_usage_models(
         entries=rows,