diff --git a/src/transformers/models/llama4/modeling_llama4.py b/src/transformers/models/llama4/modeling_llama4.py
index 6860cf30d884..032100e7a6e3 100644
--- a/src/transformers/models/llama4/modeling_llama4.py
+++ b/src/transformers/models/llama4/modeling_llama4.py
@@ -1186,7 +1186,10 @@ def __init__(self, config: Llama4Config):
         self.multi_modal_projector = Llama4MultiModalProjector(config)
         self.language_model = Llama4ForCausalLM(config.text_config)
         self.vocab_size = config.text_config.vocab_size
-        self.pad_token_id = self.config.pad_token_id if self.config.pad_token_id is not None else -1
+        if hasattr(self.config, "pad_token_id"):
+            self.pad_token_id = self.config.pad_token_id
+        else:
+            self.pad_token_id = self.config.text_config.pad_token_id or -1
 
         self.post_init()