diff --git a/src/lighteval/tasks/tasks/mmlu_pro.py b/src/lighteval/tasks/tasks/mmlu_pro.py
index 549f957be..b0beb3124 100644
--- a/src/lighteval/tasks/tasks/mmlu_pro.py
+++ b/src/lighteval/tasks/tasks/mmlu_pro.py
@@ -36,7 +36,7 @@
 
 
 TEMPLATE = """
-Answer the following multiple choice question. The last line of your response should be of the following format: 'Answer: $LETTER' (without quotes) where LETTER is one of ABCD. Think step by step before answering.
+Answer the following multiple choice question. The last line of your response should be of the following format: 'Answer: $LETTER' (without quotes) where LETTER is one of {letters}. Think step by step before answering.
 
 {question}
 
@@ -46,9 +46,11 @@
 
 
 def mmlu_pro_prompt_function(line, task_name: str = None):
-    choices = "\n".join([f"{letter}: {choice}" for letter, choice in zip(ascii_uppercase, line["options"])])
+    letters = list(ascii_uppercase)[: len(line["options"])]
+    choices = "\n".join([f"{letter}: {choice}" for letter, choice in zip(letters, line["options"])])
 
     query = TEMPLATE.format(
+        letters="".join(letters),
         question=line["question"],
         choices=choices,
     )
@@ -56,7 +58,7 @@ def mmlu_pro_prompt_function(line, task_name: str = None):
     return Doc(
         task_name=task_name,
         query=query,
-        choices=ascii_uppercase[: len(choices)],
+        choices=letters,
         gold_index=line["answer_index"],
         instruction=query,
     )
diff --git a/tests/unit/tasks/test_mmlu_pro.py b/tests/unit/tasks/test_mmlu_pro.py
new file mode 100644
index 000000000..c549eadcd
--- /dev/null
+++ b/tests/unit/tasks/test_mmlu_pro.py
@@ -0,0 +1,70 @@
+# MIT License
+
+# Copyright (c) 2024 The HuggingFace Team
+
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+
+from lighteval.tasks.tasks.mmlu_pro import mmlu_pro_prompt_function
+
+
+def _line(num_options, answer_index=0, question="What is 2 + 2?"):
+    return {
+        "question": question,
+        "options": [f"option {i}" for i in range(num_options)],
+        "answer_index": answer_index,
+    }
+
+
+def test_mmlu_pro_choices_track_option_count():
+    # MMLU-Pro questions have up to 10 options, so choices must be one letter per
+    # option. Regression for choices=ascii_uppercase[: len(choices)], where
+    # `choices` was the joined prompt string, returning all 26 letters A-Z.
+    doc = mmlu_pro_prompt_function(_line(10, answer_index=3), task_name="mmlu_pro")
+    assert doc.choices == list("ABCDEFGHIJ")
+    assert len(doc.choices) == 10
+
+
+def test_mmlu_pro_instruction_lists_real_letters():
+    # The instruction must enumerate the actual answer letters, not a hardcoded
+    # "ABCD", otherwise the model is told only A-D are valid for a 10-option
+    # question.
+    doc = mmlu_pro_prompt_function(_line(10), task_name="mmlu_pro")
+    assert "one of ABCDEFGHIJ" in doc.query
+    assert "one of ABCD." not in doc.query
+
+
+def test_mmlu_pro_four_options():
+    # Letters track the option count for shorter questions too.
+    doc = mmlu_pro_prompt_function(_line(4), task_name="mmlu_pro")
+    assert doc.choices == list("ABCD")
+    assert "one of ABCD." in doc.query
+
+
+def test_mmlu_pro_gold_index_letter_alignment():
+    # gold_index still points at the correct answer letter.
+    doc = mmlu_pro_prompt_function(_line(10, answer_index=3), task_name="mmlu_pro")
+    assert doc.gold_index == 3
+    assert doc.choices[doc.gold_index] == "D"
+
+
+def test_mmlu_pro_choices_block_enumerates_each_option():
+    doc = mmlu_pro_prompt_function(_line(10), task_name="mmlu_pro")
+    for letter in "ABCDEFGHIJ":
+        assert f"{letter}: " in doc.query