diff --git a/requirements.txt b/requirements.txt
index bc02276..eaa9b0c 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -9,3 +9,4 @@ transformers
 accelerate
 pandas
 pandas-stubs
+lm-eval>=0.4.2
diff --git a/src/instructlab/eval/mmlu.py b/src/instructlab/eval/mmlu.py
index edeea0f..b95f476 100644
--- a/src/instructlab/eval/mmlu.py
+++ b/src/instructlab/eval/mmlu.py
@@ -1,24 +1,39 @@
 # SPDX-License-Identifier: Apache-2.0
 
-# Local
-from .evaluator import Evaluator
+# Standard
+import os
 
+# Third Party
+from lm_eval.evaluator import simple_evaluate  # type: ignore
 
-class MMLU_Evaluator(Evaluator):
+# First Party
+from instructlab.eval.evaluator import Evaluator
+
+
+class MMLUEvaluator(Evaluator):
     """
     Child class of an Evaluator for Massive Multitask Language Understanding (MMLU)
 
     Attributes:
+        model_path   absolute path to or name of a huggingface model
         tasks        list of tasks for MMLU to test the model with
+        model_dtype  dtype of model when served
         few_shots    number of examples
         batch_size   number of GPUs
     """
 
     def __init__(
-        self, model_path, tasks: list[str], few_shots: int = 2, batch_size: int = 5
+        self,
+        model_path,
+        tasks: list[str],
+        model_dtype="bfloat16",
+        few_shots: int = 2,
+        batch_size: int = 5,
     ) -> None:
+        super().__init__()
         self.model_path = model_path
         self.tasks = tasks
+        self.model_dtype = model_dtype
         self.few_shots = few_shots
         self.batch_size = batch_size
 
@@ -30,18 +45,42 @@ def run(self) -> tuple:
             overall_score       MMLU score for the overall model evaluation
             individual_scores   Individual MMLU score for each task
         """
-        individual_scores: dict[str, float] = {}
-        overall_score: float = 0.0
+        # TODO: make this a parameter for class?
+        os.environ["TOKENIZERS_PARALLELISM"] = "true"
+
+        individual_scores: dict = {}
+        agg_score: float = 0.0
+        model_args = f"pretrained= {self.model_path}, dtype= {self.model_dtype}"
+
+        mmlu_output = simple_evaluate(
+            model="hf",
+            model_args=model_args,
+            tasks=self.tasks,
+            num_fewshot=self.few_shots,
+            batch_size=self.batch_size,
+        )
+
+        results = mmlu_output["results"]
+
+        for task in self.tasks:
+            mmlu_res = results[task]
+            agg_score += float(mmlu_res["acc,none"])
+            individual_scores[task] = {}
+            individual_scores[task]["score"] = float(mmlu_res["acc,none"])
+            individual_scores[task]["stderr"] = float(mmlu_res["acc_stderr,none"])
+
+        overall_score = float(agg_score / len(self.tasks))
         return overall_score, individual_scores
 
 
-class PR_MMLU_Evaluator(Evaluator):
+class MMLUBranchEvaluator(Evaluator):
     """
-    Child class of an Evaluator for PR Massive Multitask Language Understanding (PR MMLU)
+    Child class of an Evaluator for Massive Multitask Language Understanding Branch (MMLUBranch)
 
     Attributes:
-        sdg_path    path where all the PR MMLU tasks are stored
-        task        group name that is shared by all the PR MMLU tasks
+        model_path  absolute path to or name of a huggingface model
+        sdg_path    path where all the MMLUBranch tasks are stored
+        task        group name that is shared by all the MMLUBranch tasks
         few_shots   number of examples
         batch_size  number of GPUs
     """
@@ -62,11 +101,11 @@ def __init__(
 
     def run(self) -> tuple:
         """
-        Runs PR MMLU evaluation
+        Runs MMLUBranch evaluation
 
         Returns:
-            overall_score       PR MMLU score for the overall model evaluation
-            individual_scores   Individual PR MMLU scores for each task
+            overall_score       MMLUBranch score for the overall model evaluation
+            individual_scores   Individual MMLUBranch scores for each task
             qa_pairs            Question and answer pairs from the evaluation
         """
         individual_scores: dict[str, float] = {}