diff --git a/vllm/model_executor/models/chameleon.py b/vllm/model_executor/models/chameleon.py
index 7659f598bab94..10a82207d90ef 100644
--- a/vllm/model_executor/models/chameleon.py
+++ b/vllm/model_executor/models/chameleon.py
@@ -998,6 +998,13 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
                 # Models trained using ColossalAI may include these tensors in
                 # the checkpoint. Skip them.
                 continue
+
+            # With tie_word_embeddings, we can skip lm_head.weight
+            # The weight might appear unnecessarily in the files if the model is
+            # processed with quantization, LoRA, fine-tuning, etc.
+            if self.config.tie_word_embeddings and "lm_head.weight" in name:
+                continue
+
             use_default_weight_loading = False
             if "vqmodel" in name:
                 if self.model.vqmodel is not None:
diff --git a/vllm/model_executor/models/llama.py b/vllm/model_executor/models/llama.py
index 2052c443a8885..048c292c79c83 100644
--- a/vllm/model_executor/models/llama.py
+++ b/vllm/model_executor/models/llama.py
@@ -469,6 +469,11 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
                 # Models trained using ColossalAI may include these tensors in
                 # the checkpoint. Skip them.
                 continue
+            # With tie_word_embeddings, we can skip lm_head.weight
+            # The weight might appear unnecessarily in the files if the model is
+            # processed with quantization, LoRA, fine-tuning, etc.
+            if self.config.tie_word_embeddings and "lm_head.weight" in name:
+                continue
             if scale_name := get_compressed_tensors_cache_scale(name):
                 # Loading kv cache scales for compressed-tensors quantization
                 param = params_dict[scale_name]
diff --git a/vllm/model_executor/models/minicpm.py b/vllm/model_executor/models/minicpm.py
index b46e88f5fc584..7f8f38fe8439a 100644
--- a/vllm/model_executor/models/minicpm.py
+++ b/vllm/model_executor/models/minicpm.py
@@ -514,7 +514,11 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
                 # Models trained using ColossalAI may include these tensors in
                 # the checkpoint. Skip them.
                 continue
-
+            # With tie_word_embeddings, we can skip lm_head.weight
+            # The weight might appear unnecessarily in the files if the model is
+            # processed with quantization, LoRA, fine-tuning, etc.
+            if self.config.tie_word_embeddings and "lm_head.weight" in name:
+                continue
             for (param_name, weight_name, shard_id) in stacked_params_mapping:
                 if weight_name not in name:
                     continue
diff --git a/vllm/model_executor/models/olmo.py b/vllm/model_executor/models/olmo.py
index 408c0c883a9d0..1a0a3774dc8fb 100644
--- a/vllm/model_executor/models/olmo.py
+++ b/vllm/model_executor/models/olmo.py
@@ -343,6 +343,11 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
                 # Models trained using ColossalAI may include these tensors in
                 # the checkpoint. Skip them.
                 continue
+            # With tie_word_embeddings, we can skip lm_head.weight
+            # The weight might appear unnecessarily in the files if the model is
+            # processed with quantization, LoRA, fine-tuning, etc.
+            if self.config.tie_word_embeddings and "lm_head.weight" in name:
+                continue
             for (param_name, weight_name, shard_id) in stacked_params_mapping:
                 if weight_name not in name:
                     continue