diff --git a/promptbench/models/models.py b/promptbench/models/models.py
index 4a23344..adc81ed 100644
--- a/promptbench/models/models.py
+++ b/promptbench/models/models.py
@@ -268,6 +268,7 @@ def __init__(self, model_name, max_new_tokens, temperature=0, system_prompt=None
         super(OpenAIModel, self).__init__(model_name, max_new_tokens, temperature)
         self.openai_key = openai_key
         self.sleep_time = sleep_time
+        self.system_prompt = system_prompt
 
         if self.temperature > 0:
             raise Warning("Temperature is not 0, so that the results may not be reproducable!")
@@ -283,8 +284,11 @@ def predict(self, input_text, **kwargs):
         
         from openai import OpenAI
         client = OpenAI(api_key=self.openai_key)
-        
-        system_messages = {'role': "system", 'content': "You are a helpful assistant."}
+
+        if self.system_prompt is None:
+            system_messages = {'role': "system", 'content': "You are a helpful assistant."}
+        else:
+            system_messages = {'role': "system", 'content': self.system_prompt}
         
         if isinstance(input_text, list):
             messages = input_text
@@ -346,15 +350,15 @@ class PaLMModel(LMMBaseModel):
     model_dir : str, optional
         The directory containing the model files (default is None).
     """
-    def __init__(self, model, max_new_tokens, temperature=0, system_prompt=None, palm_key=None, sleep_time=3):
+    def __init__(self, model, max_new_tokens, temperature=0, system_prompt=None, api_key=None, sleep_time=3):
         super(PaLMModel, self).__init__(model, max_new_tokens, temperature)
-        self.palm_key = palm_key
+        self.api_key = api_key
         self.sleep_time = sleep_time
     
     def predict(self, input_text, **kwargs):
         import google.generativeai as palm 
         
-        palm.configure(api_key=self.palm_key)
+        palm.configure(api_key=self.api_key)
         models = [m for m in palm.list_models() if 'generateText' in m.supported_generation_methods]
         model = models[0].name
         
@@ -377,7 +381,6 @@ def predict(self, input_text, **kwargs):
         
         return result
         
-
 class GeminiModel(LMMBaseModel):
     """
     Language model class for interfacing with Google's Gemini models.