diff --git a/src/transformers/models/qwen3_moe/modeling_qwen3_moe.py b/src/transformers/models/qwen3_moe/modeling_qwen3_moe.py
index 77b2362fe19..d4ba6c36569 100644
--- a/src/transformers/models/qwen3_moe/modeling_qwen3_moe.py
+++ b/src/transformers/models/qwen3_moe/modeling_qwen3_moe.py
@@ -355,6 +355,7 @@ class Qwen3MoeDecoderLayer(nn.Module):
             use_cache=use_cache,
             cache_position=cache_position,
             position_embeddings=position_embeddings,
+            **kwargs,
         )
         hidden_states = residual + hidden_states
 
diff --git a/src/transformers/models/qwen3_moe/modular_qwen3_moe.py b/src/transformers/models/qwen3_moe/modular_qwen3_moe.py
index 7170164f515..6f821d5b814 100644
--- a/src/transformers/models/qwen3_moe/modular_qwen3_moe.py
+++ b/src/transformers/models/qwen3_moe/modular_qwen3_moe.py
@@ -189,6 +189,7 @@ class Qwen3MoeDecoderLayer(Qwen2MoeDecoderLayer, nn.Module):
             use_cache=use_cache,
             cache_position=cache_position,
             position_embeddings=position_embeddings,
+            **kwargs,
         )
         hidden_states = residual + hidden_states