diff --git a/tests/test_full_model.py b/tests/test_full_model.py
index 29be828..1c68051 100644
--- a/tests/test_full_model.py
+++ b/tests/test_full_model.py
@@ -142,8 +142,8 @@ def test_sampling(tokenizer, model, ref_model, max_new_tokens=10):
 
 
 @pytest.mark.skipif(
-    MODEL_NAME.lower().find("mixtral") > -1,
-    reason="Mixtral uses DynamicCache, which can change based on beam search choices",
+    MODEL_NAME.lower().find("bloom") == -1,
+    reason="Mixtral and Llama uses DynamicCache, which can change based on beam search choices",
 )
 @pytest.mark.forked
 def test_beam_search_generation(tokenizer, model, ref_model, max_new_tokens=4, num_beams=5):