diff --git a/README.md b/README.md
index 7c8d667..442fc7f 100644
--- a/README.md
+++ b/README.md
@@ -21,7 +21,7 @@ scripts/infra/cloud-instance.sh ec2 ssh
 # Regardless of how you setup your instance
 git clone https://github.com/instructlab/taxonomy.git && pushd taxonomy && git branch rc && popd
 git clone --bare https://github.com/instructlab/eval.git && git clone eval.git/ && cd eval && git remote add syncrepo ../eval.git
-python -m venv venv
+python3 -m venv venv
 source venv/bin/activate
 pip install -r requirements.txt
 pip install -r requirements-dev.txt
@@ -104,4 +104,4 @@ eval_output/
         └── reference_answer
             └── instructlab
                 └── granite-7b-lab.jsonl
-```
\ No newline at end of file
+```
diff --git a/src/instructlab/eval/exceptions.py b/src/instructlab/eval/exceptions.py
index 65a82f3..65b4910 100644
--- a/src/instructlab/eval/exceptions.py
+++ b/src/instructlab/eval/exceptions.py
@@ -90,3 +90,16 @@ def __init__(self, tasks_dir) -> None:
         super().__init__()
         self.tasks_dir = tasks_dir
         self.message = f"Invalid Tasks Dir: {tasks_dir}"
+
+
+class CompletionError(EvalError):
+    """
+    Error raised when the model failed to provide a completion.
+    Attributes
+        message         error message to be printed on raise
+        tasks_dir       tasks dir
+    """
+
+    def __init__(self) -> None:
+        super().__init__()
+        self.message = "Failed to retrieve completion."
diff --git a/src/instructlab/eval/mt_bench_common.py b/src/instructlab/eval/mt_bench_common.py
index d346999..3956154 100644
--- a/src/instructlab/eval/mt_bench_common.py
+++ b/src/instructlab/eval/mt_bench_common.py
@@ -17,6 +17,9 @@
 from fastchat.model.model_adapter import get_conversation_template  # type: ignore
 import openai
 
+# First Party
+from instructlab.eval import exceptions
+
 # Local
 from .logger_config import setup_logger
 
@@ -250,9 +253,10 @@ def play_a_match_single(
 def chat_completion_openai(
     openai_client, model, conv, temperature, max_tokens, merge_system_user_message=False
 ) -> str:
-    output = API_ERROR_OUTPUT
     for i in range(API_MAX_RETRY):
         try:
+            if i > 0:
+                time.sleep(API_RETRY_SLEEP)
             messages = conv.to_openai_api_messages()
             if (
                 merge_system_user_message
@@ -270,17 +274,16 @@ def chat_completion_openai(
                 temperature=temperature,
                 max_tokens=max_tokens,
             )
-            output = response.choices[0].message.content
-            break
-        except openai.OpenAIError as e:
-            if i == API_MAX_RETRY - 1:
-                # Print error on last try
-                print(type(e), e)
-            else:
-                logger.debug(e)
-            time.sleep(API_RETRY_SLEEP)
-
-    return output
+            return response.choices[0].message.content
+        except openai.APITimeoutError as ex:
+            logger.debug(ex)
+        except openai.APIConnectionError as ex:
+            logger.debug(ex)
+            raise exceptions.EvalError("Failed to connect to API.")
+        except openai.OpenAIError as ex:
+            logger.debug(ex)
+
+    raise exceptions.EvalError("Failed to get completion from API.")
 
 
 def check_data(questions, model_answers, ref_answers, models, judges):