diff --git a/src/pyob/core_utils.py b/src/pyob/core_utils.py
index b2581b0..9ee013d 100644
--- a/src/pyob/core_utils.py
+++ b/src/pyob/core_utils.py
@@ -487,10 +487,12 @@ def get_valid_llm_response(self, prompt: str, validator, context: str = "") -> s
 
             # 3. Handle Empty or Error Responses (STOPS THE INFINITE LOOP)
             if not response_text or response_text.startswith("ERROR_CODE_"):
+                # If we are in the cloud, we need to wait longer for the token bucket to refill
+                wait_time = 60 if is_cloud else 10
                 logger.warning(
-                    "⚠️ API Error/Empty Response. Sleeping 10s before retry..."
+                    f"⚠️ API Error/Empty Response. Sleeping {wait_time}s to refill tokens..."
                 )
-                time.sleep(10)  # MANDATORY SLEEP to prevent tight-looping
+                time.sleep(wait_time)
                 attempts += 1
                 continue