diff --git a/benchmarks/benchmark_throughput.py b/benchmarks/benchmark_throughput.py
index 6df1e1d628e6c..79d7e9adc4515 100644
--- a/benchmarks/benchmark_throughput.py
+++ b/benchmarks/benchmark_throughput.py
@@ -335,7 +335,7 @@ if __name__ == "__main__":
         "--device",
         type=str,
         default="cuda",
-        choices=["cuda", "cpu"],
+        choices=["cuda", "cpu", "tpu"],
         help='device type for vLLM execution, supporting CUDA and CPU.')
     parser.add_argument(
         "--enable-prefix-caching",