diff --git a/chromadb/utils/batch_utils.py b/chromadb/utils/batch_utils.py
index f3881bc6a03..9fa210cbcfb 100644
--- a/chromadb/utils/batch_utils.py
+++ b/chromadb/utils/batch_utils.py
@@ -1,3 +1,4 @@
+import random
 from typing import Optional, Tuple, List
 from chromadb.api import BaseAPI
 from chromadb.api.types import (
@@ -18,19 +19,22 @@ def create_batches(
     _batches: List[
         Tuple[IDs, Embeddings, Optional[Metadatas], Optional[Documents]]
     ] = []
-    if len(ids) > api.get_max_batch_size():
-        # create split batches
-        for i in range(0, len(ids), api.get_max_batch_size()):
+    max_batch_size = api.get_max_batch_size()
+    offset = 0
+    if len(ids) > max_batch_size:
+        while offset < len(ids):
+            batch_size = random.randint(1, max_batch_size)
             _batches.append(
                 (  # type: ignore
-                    ids[i : i + api.get_max_batch_size()],
-                    embeddings[i : i + api.get_max_batch_size()]
+                    ids[offset : offset + batch_size],
+                    embeddings[offset : offset + batch_size]
                     if embeddings
                     else None,
-                    metadatas[i : i + api.get_max_batch_size()] if metadatas else None,
-                    documents[i : i + api.get_max_batch_size()] if documents else None,
+                    metadatas[offset : offset + batch_size] if metadatas else None,
+                    documents[offset : offset + batch_size] if documents else None,
                 )
             )
+            offset += batch_size
     else:
         _batches.append((ids, embeddings, metadatas, documents))  # type: ignore
     return _batches