diff --git a/vllm/v1/worker/gpu/async_utils.py b/vllm/v1/worker/gpu/async_utils.py
index f87459efa..7f270c2b8 100644
--- a/vllm/v1/worker/gpu/async_utils.py
+++ b/vllm/v1/worker/gpu/async_utils.py
@@ -95,8 +95,8 @@ class AsyncPoolingOutput(AsyncModelRunnerOutput):
             self.copy_event.record(copy_stream)
 
     def get_output(self) -> ModelRunnerOutput:
+        pooler_output = list(self.pooler_output_cpu.unbind(dim=0))
         self.copy_event.synchronize()
-        pooler_output = self.pooler_output_cpu.unbind(dim=0)
         if self.is_valid_cpu is not None:
             is_valid_cpu = self.is_valid_cpu.tolist()
             for i, is_valid in enumerate(is_valid_cpu):
diff --git a/vllm/v1/worker/gpu/model_runner.py b/vllm/v1/worker/gpu/model_runner.py
index 203d31195..9f802ed76 100644
--- a/vllm/v1/worker/gpu/model_runner.py
+++ b/vllm/v1/worker/gpu/model_runner.py
@@ -1117,7 +1117,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
             # The prior execute_model call must have failed.
             return None
 
-        input_batch, _, _, _, hidden_states, _, kv_connector_output = (
+        input_batch, _, _, _, hidden_states, _, kv_connector_output, _ = (
             self.execute_model_state
         )
         self.execute_model_state = None