diff --git a/vllm/v1/engine/async_llm.py b/vllm/v1/engine/async_llm.py
index fc57fed20..380a1567b 100644
--- a/vllm/v1/engine/async_llm.py
+++ b/vllm/v1/engine/async_llm.py
@@ -9,7 +9,6 @@ from collections.abc import AsyncGenerator, Iterable, Mapping
 from copy import copy
 from typing import Any, cast
 
-import numpy as np
 import torch
 
 import vllm.envs as envs
@@ -32,7 +31,6 @@ from vllm.transformers_utils.config import maybe_register_config_serialize_by_va
 from vllm.usage.usage_lib import UsageContext
 from vllm.utils.async_utils import cancel_task_threadsafe
 from vllm.utils.collection_utils import as_list
-from vllm.utils.math_utils import cdiv
 from vllm.v1.engine import EngineCoreRequest
 from vllm.v1.engine.core_client import EngineCoreClient
 from vllm.v1.engine.exceptions import EngineDeadError, EngineGenerateError
@@ -495,6 +493,7 @@ class AsyncLLM(EngineClient):
         log_stats = self.log_stats
         logger_manager = self.logger_manager
         input_processor = self.input_processor
+        chunk_size = envs.VLLM_V1_OUTPUT_PROC_CHUNK_SIZE
 
         async def output_handler():
             try:
@@ -510,15 +509,10 @@ class AsyncLLM(EngineClient):
                     # Split outputs into chunks of at most
                     # VLLM_V1_OUTPUT_PROC_CHUNK_SIZE, so that we don't block the
                     # event loop for too long.
-                    if num_outputs <= envs.VLLM_V1_OUTPUT_PROC_CHUNK_SIZE:
-                        slices = (outputs.outputs,)
-                    else:
-                        slices = np.array_split(
-                            outputs.outputs,
-                            cdiv(num_outputs, envs.VLLM_V1_OUTPUT_PROC_CHUNK_SIZE),
-                        )
-
-                    for i, outputs_slice in enumerate(slices):
+                    engine_core_outputs = outputs.outputs
+                    for start in range(0, num_outputs, chunk_size):
+                        end = start + chunk_size
+                        outputs_slice = engine_core_outputs[start:end]
                         # 2) Process EngineCoreOutputs.
                         processed_outputs = output_processor.process_outputs(
                             outputs_slice, outputs.timestamp, iteration_stats
@@ -527,7 +521,7 @@ class AsyncLLM(EngineClient):
                         assert not processed_outputs.request_outputs
 
                         # Allow other asyncio tasks to run between chunks
-                        if i + 1 < len(slices):
+                        if end < num_outputs:
                             await asyncio.sleep(0)
 
                         # 3) Abort any reqs that finished due to stop strings.