diff --git a/gpt4all-backend/llamamodel.cpp b/gpt4all-backend/llamamodel.cpp
index 5ab89e2b..7c66be9f 100644
--- a/gpt4all-backend/llamamodel.cpp
+++ b/gpt4all-backend/llamamodel.cpp
@@ -940,6 +940,8 @@ void LLamaModel::embedInternal(
     }
 
     if (tokenCount) { *tokenCount = totalTokens; }
+
+    llama_batch_free(batch);
 }
 
 #if defined(_WIN32)