diff --git a/python/sglang/srt/models/qwen2.py b/python/sglang/srt/models/qwen2.py
index 3876d6444..98a3b20d9 100644
--- a/python/sglang/srt/models/qwen2.py
+++ b/python/sglang/srt/models/qwen2.py
@@ -462,7 +462,7 @@ class Qwen2ForCausalLM(nn.Module):
                 self.pp_group.send(
                     self.model.embed_tokens.weight, dst=self.pp_group.last_rank
                 )
-            else:
+            elif self.pp_group.is_last_rank:
                 emb_token_weight = self.pp_group.recv(
                     size=(config.vocab_size, config.hidden_size),
                     dtype=next(self.model.parameters()).dtype,
diff --git a/python/sglang/srt/models/qwen3.py b/python/sglang/srt/models/qwen3.py
index b1135a6b2..b031d6e03 100644
--- a/python/sglang/srt/models/qwen3.py
+++ b/python/sglang/srt/models/qwen3.py
@@ -361,7 +361,7 @@ class Qwen3ForCausalLM(nn.Module):
                 self.pp_group.send(
                     self.model.embed_tokens.weight, dst=self.pp_group.last_rank
                 )
-            else:
+            elif self.pp_group.is_last_rank:
                 emb_token_weight = self.pp_group.recv(
                     size=(config.vocab_size, config.hidden_size),
                     dtype=next(self.model.parameters()).dtype,