diff --git a/libs/vertexai/langchain_google_vertexai/embeddings.py b/libs/vertexai/langchain_google_vertexai/embeddings.py
index df97a357..1e42e4be 100644
--- a/libs/vertexai/langchain_google_vertexai/embeddings.py
+++ b/libs/vertexai/langchain_google_vertexai/embeddings.py
@@ -320,7 +320,6 @@ def _prepare_and_validate_batches(
                     first_result = self._get_embeddings_with_retry(
                         first_batch, embeddings_type
                     )
-                    batches = batches[1:]
                     break
                 except InvalidArgument:
                     had_failure = True
@@ -347,6 +346,8 @@ def _prepare_and_validate_batches(
                     batches = VertexAIEmbeddings._prepare_batches(
                         texts[first_batch_len:], self.instance["batch_size"]
                     )
+                else:
+                    batches = batches[1:]
             else:
                 # Still figuring out max batch size.
                 batches = batches[1:]
diff --git a/libs/vertexai/tests/unit_tests/test_embeddings.py b/libs/vertexai/tests/unit_tests/test_embeddings.py
index efba09d1..287fbb31 100644
--- a/libs/vertexai/tests/unit_tests/test_embeddings.py
+++ b/libs/vertexai/tests/unit_tests/test_embeddings.py
@@ -19,7 +19,7 @@ def test_langchain_google_vertexai_embed_image_multimodal_only() -> None:
 def test_langchain_google_vertexai_no_dups_dynamic_batch_size() -> None:
     mock_embeddings = MockVertexAIEmbeddings("textembedding-gecko@001")
     default_batch_size = mock_embeddings.instance["batch_size"]
-    texts = ["text_{i}" for i in range(default_batch_size * 2)]
+    texts = ["text {i}" for i in range(default_batch_size * 2)]
     # It should only return one batch (out of two) still to process
     _, batches = mock_embeddings._prepare_and_validate_batches(texts=texts)
     assert len(batches) == 1