diff --git a/litellm/llms/mistral/audio_speech/transformation.py b/litellm/llms/mistral/audio_speech/transformation.py
index 6d1a693268a..e7f7d510346 100644
--- a/litellm/llms/mistral/audio_speech/transformation.py
+++ b/litellm/llms/mistral/audio_speech/transformation.py
@@ -172,7 +172,7 @@ class MistralTextToSpeechConfig(BaseTextToSpeechConfig):
                 headers=raw_response.headers,
             )
         try:
-            audio_bytes: Final = base64.b64decode(audio_b64)
+            audio_bytes: Final = base64.b64decode(audio_b64, validate=True)
         except ValueError:
             raise MistralTextToSpeechException(
                 status_code=500,
diff --git a/tests/test_litellm/llms/mistral/audio_speech/test_mistral_audio_speech_transformation.py b/tests/test_litellm/llms/mistral/audio_speech/test_mistral_audio_speech_transformation.py
index 20d07699cb8..6d250901e50 100644
--- a/tests/test_litellm/llms/mistral/audio_speech/test_mistral_audio_speech_transformation.py
+++ b/tests/test_litellm/llms/mistral/audio_speech/test_mistral_audio_speech_transformation.py
@@ -187,7 +187,7 @@ def test_transform_response_invalid_base64_raises():
     config: Final = MistralTextToSpeechConfig()
     raw_response: Final = httpx.Response(
         status_code=200,
-        json={"audio_data": "!!!not-base64!!!"},
+        json={"audio_data": "QUJD!QUJD"},
         request=httpx.Request("POST", SPEECH_URL),
     )
     with pytest.raises(MistralTextToSpeechException, match="base64"):