diff --git a/tests/entrypoints/serve/utils/test_request_logger.py b/tests/entrypoints/serve/utils/test_request_logger.py
index 461b954cae0f..065bc012f91f 100644
--- a/tests/entrypoints/serve/utils/test_request_logger.py
+++ b/tests/entrypoints/serve/utils/test_request_logger.py
@@ -284,6 +284,7 @@ def test_request_logger_log_request_body_truncation(max_log_len):
     with patch("vllm.entrypoints.serve.utils.request_logger.logger", mock_logger):
         RequestLogger(max_log_len=max_log_len).log_request_body(request)
 
-    mock_logger.debug.assert_called_once()
-    logged_body = mock_logger.debug.call_args.args[2]
+    mock_logger.log.assert_called_once()
+    assert mock_logger.log.call_args.args[0] == logging.DEBUG
+    logged_body = mock_logger.log.call_args.args[3]
     assert logged_body == full_body[:max_log_len]
diff --git a/vllm/engine/arg_utils.py b/vllm/engine/arg_utils.py
index 8f285317e04b..ccf66f15b9dc 100644
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -3242,7 +3242,9 @@ def add_cli_args(
             help="Enable logging request information, dependent on log level:\n"
             "- INFO: Request ID, parameters and LoRA request.\n"
             "- DEBUG: Prompt inputs (e.g: text, token IDs).\n"
-            "You can set the minimum log level via `VLLM_LOGGING_LEVEL`.",
+            "You can set the minimum log level via `VLLM_LOGGING_LEVEL`.\n"
+            "You can also set a separate log level for request JSON body "
+            "via `VLLM_REQUEST_BODY_LOG_LEVEL` (default: DEBUG).",
         )
         current_platform.pre_register_and_update(parser)
         return parser
diff --git a/vllm/entrypoints/serve/utils/request_logger.py b/vllm/entrypoints/serve/utils/request_logger.py
index 6272054a1e99..412a2d40e134 100644
--- a/vllm/entrypoints/serve/utils/request_logger.py
+++ b/vllm/entrypoints/serve/utils/request_logger.py
@@ -6,6 +6,7 @@
 
 import torch
 
+from vllm import envs
 from vllm.entrypoints.pooling.typing import AnyPoolingRequest
 from vllm.entrypoints.serve.engine.typing import AnyRequest
 from vllm.logger import init_logger
@@ -70,12 +71,22 @@ def log_inputs(
         )
 
     def log_request_body(self, request: AnyRequest | AnyPoolingRequest) -> None:
-        if logger.isEnabledFor(logging.DEBUG):
-            logger.debug(
-                "Request %s JSON body: %s",
-                getattr(request, "request_id", "N/A"),
-                request.model_dump_json(exclude_unset=True)[: self.max_log_len],
+        try:
+            level = logging.getLevelNamesMapping()[envs.VLLM_REQUEST_BODY_LOG_LEVEL]
+        except KeyError:
+            logger.warning_once(
+                "Invalid logging level for request content: %s. "
+                "Request JSON body will not be logged.",
+                envs.VLLM_REQUEST_BODY_LOG_LEVEL,
             )
+            return
+
+        logger.log(
+            level,
+            "Request %s JSON body: %s",
+            getattr(request, "request_id", "N/A"),
+            request.model_dump_json(exclude_unset=True)[: self.max_log_len],
+        )
 
     def log_outputs(
         self,
diff --git a/vllm/envs.py b/vllm/envs.py
index 13922923ed57..19e3250e2fbe 100755
--- a/vllm/envs.py
+++ b/vllm/envs.py
@@ -331,6 +331,7 @@
     VLLM_GPU_NIC_PCIE_MAPPING: str = ""
     VLLM_NIC_SELECTION_VARS: str = ""
     VLLM_ENABLE_HPC_OPS: bool = False
+    VLLM_REQUEST_BODY_LOG_LEVEL: str = "DEBUG"
 
 
 def get_default_cache_root():
@@ -2264,6 +2265,10 @@ def _resolve_rust_cli_path() -> str | None:
     "VLLM_SKIP_VERSION_SUFFIX": lambda: bool(
         int(os.getenv("VLLM_SKIP_VERSION_SUFFIX", "0"))
     ),
+    # Logging level for request JSON body.
+    "VLLM_REQUEST_BODY_LOG_LEVEL": lambda: os.getenv(
+        "VLLM_REQUEST_BODY_LOG_LEVEL", "DEBUG"
+    ).upper(),
 }
 
 
@@ -2423,6 +2428,7 @@ def compile_factors() -> dict[str, object]:
         "NO_COLOR",
         "VLLM_SKIP_VERSION_SUFFIX",
         "FORCE_COLOR",
+        "VLLM_REQUEST_BODY_LOG_LEVEL",
     }
 
     from vllm.config.utils import normalize_value