diff --git a/examples/vllm_serve/fakequant_worker.py b/examples/vllm_serve/fakequant_worker.py
index 4f84df0581..55c4e3398a 100644
--- a/examples/vllm_serve/fakequant_worker.py
+++ b/examples/vllm_serve/fakequant_worker.py
@@ -20,7 +20,6 @@
 
 import torch
 from transformers import AutoTokenizer
-from vllm.v1.worker.gpu_worker import Worker as BaseWorker
 from vllm_ptq_utils import calibrate_fun, get_quant_config
 from vllm_reload_utils import (
     convert_dict_to_vllm,
@@ -38,6 +37,7 @@
 )
 from modelopt.torch.utils import safe_load
 from modelopt.torch.utils.dataset_utils import get_dataset_dataloader
+from vllm.v1.worker.gpu_worker import Worker as BaseWorker
 
 quant_config: dict[str, Any] = {
     "dataset": os.environ.get("QUANT_DATASET", "cnn_dailymail"),
diff --git a/examples/vllm_serve/vllm_ptq_utils.py b/examples/vllm_serve/vllm_ptq_utils.py
index 88b31d54a7..55616df314 100644
--- a/examples/vllm_serve/vllm_ptq_utils.py
+++ b/examples/vllm_serve/vllm_ptq_utils.py
@@ -20,11 +20,11 @@
 import torch
 from torch.utils.data import DataLoader
 from tqdm import tqdm
-from vllm.sampling_params import SamplingParams
-from vllm.v1.core.sched.output import CachedRequestData, NewRequestData, SchedulerOutput
 
 import modelopt.torch.quantization as mtq
 from modelopt.recipe import ModelOptPTQRecipe, load_recipe
+from vllm.sampling_params import SamplingParams
+from vllm.v1.core.sched.output import CachedRequestData, NewRequestData, SchedulerOutput
 
 
 def _create_new_data_cls(data_cls, **kwargs):
diff --git a/examples/vllm_serve/vllm_reload_utils.py b/examples/vllm_serve/vllm_reload_utils.py
index 6b658551f1..0a2e95cdf0 100644
--- a/examples/vllm_serve/vllm_reload_utils.py
+++ b/examples/vllm_serve/vllm_reload_utils.py
@@ -20,7 +20,6 @@
 from typing import Any
 
 import torch
-from vllm.distributed.parallel_state import get_tp_group
 
 from modelopt.torch.export.plugins.vllm_fakequant_hf import (
     infer_quantizer_prefix_remap,
@@ -38,6 +37,7 @@
 )
 from modelopt.torch.quantization.nn import SequentialQuantizer, TensorQuantizer
 from modelopt.torch.quantization.utils import is_quantized
+from vllm.distributed.parallel_state import get_tp_group
 
 
 def _union_quantizer_keys_across_ranks(local_quantizer_keys: list[str]) -> set[str]:
diff --git a/examples/vllm_serve/vllm_serve_fakequant.py b/examples/vllm_serve/vllm_serve_fakequant.py
index 458aaa908a..aa4ef3e4da 100644
--- a/examples/vllm_serve/vllm_serve_fakequant.py
+++ b/examples/vllm_serve/vllm_serve_fakequant.py
@@ -55,8 +55,9 @@
 from pathlib import Path
 
 import uvloop
-import vllm
 from packaging import version
+
+import vllm
 from vllm.entrypoints.openai.api_server import run_server
 from vllm.entrypoints.openai.cli_args import make_arg_parser
 
diff --git a/modelopt/torch/export/unified_export_hf.py b/modelopt/torch/export/unified_export_hf.py
index a76783ac17..ed6ed2fcf2 100644
--- a/modelopt/torch/export/unified_export_hf.py
+++ b/modelopt/torch/export/unified_export_hf.py
@@ -1189,12 +1189,25 @@ def export_hf_checkpoint(
     try:
         post_state_dict, hf_quant_config = _export_transformers_checkpoint(model, dtype)
 
-        if hf_quant_config is not None:
+        # Only treat the export as quantized when at least one quant_algo field is set.
+        # get_quant_config always returns a dict (even for sparsity-only or unmodified models),
+        # so emitting hf_quant_config.json unconditionally produces a file with
+        # "quant_algo": null that downstream loaders (e.g. TensorRT-LLM) reject as a
+        # malformed pre-quantized checkpoint.
+        quantization_details = (hf_quant_config or {}).get("quantization", {})
+        is_quantized_export = (
+            quantization_details.get("quant_algo") is not None
+            or quantization_details.get("kv_cache_quant_algo") is not None
+        )
+
+        if is_quantized_export:
             # Save hf_quant_config.json for backward compatibility
             with open(f"{export_dir}/hf_quant_config.json", "w") as file:
                 json.dump(hf_quant_config, file, indent=4)
 
             hf_quant_config = convert_hf_quant_config_format(hf_quant_config)
+        else:
+            hf_quant_config = None
 
         # Remove hf_quantizer from model so post_state_dict can be exported.
         if getattr(model, "hf_quantizer", None) is not None: