diff --git a/tensorrt_llm/_torch/models/modeling_kimi_k25.py b/tensorrt_llm/_torch/models/modeling_kimi_k25.py
index a216e916df0f..8eed050eafe2 100644
--- a/tensorrt_llm/_torch/models/modeling_kimi_k25.py
+++ b/tensorrt_llm/_torch/models/modeling_kimi_k25.py
@@ -1517,6 +1517,11 @@ class KimiK25ForConditionalGeneration(PreTrainedModel):
 
     _LANG_PREFIX = "language_model."
 
+    @classmethod
+    def get_model_defaults(cls, llm_args: Any) -> dict:
+        """Use the C++ KV cache manager V2 by default."""
+        return {"kv_cache_config": {"use_kv_cache_manager_v2": True}}
+
     @classmethod
     def get_preferred_transceiver_runtime(
         cls,