diff --git a/benchmark/hicache/data_processing.py b/benchmark/hicache/data_processing.py
index 1fb3650ce..dd0cbf669 100644
--- a/benchmark/hicache/data_processing.py
+++ b/benchmark/hicache/data_processing.py
@@ -11,10 +11,10 @@ from nextqa import NExTQALoader
 from tqdm.asyncio import tqdm
 from transformers import PreTrainedTokenizerBase
 
-SHAREGPT_URL = "https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json"
-
 from sglang.bench_serving import (
-    download_and_cache_file,
+    SHAREGPT_FILENAME,
+    SHAREGPT_REPO_ID,
+    download_and_cache_hf_file,
     gen_prompt,
     get_gen_prefix_cache_path,
 )
@@ -104,7 +104,10 @@ def sample_sharegpt_requests(
 
     # Download sharegpt if necessary
     if not os.path.isfile(dataset_path):
-        dataset_path = download_and_cache_file(SHAREGPT_URL)
+        dataset_path = download_and_cache_hf_file(
+            repo_id=SHAREGPT_REPO_ID,
+            filename=SHAREGPT_FILENAME,
+        )
 
     # Load the dataset.
     with open(dataset_path) as f:
@@ -367,7 +370,10 @@ def sample_random_requests(
 
         # Download sharegpt if necessary
         if not os.path.isfile(dataset_path):
-            dataset_path = download_and_cache_file(SHAREGPT_URL)
+            dataset_path = download_and_cache_hf_file(
+                repo_id=SHAREGPT_REPO_ID,
+                filename=SHAREGPT_FILENAME,
+            )
 
         # Load the dataset.
         with open(dataset_path) as f:
diff --git a/python/sglang/bench_serving.py b/python/sglang/bench_serving.py
index f0b3b4472..60ccb3b84 100644
--- a/python/sglang/bench_serving.py
+++ b/python/sglang/bench_serving.py
@@ -918,7 +918,8 @@ class BenchmarkMetrics:
     max_concurrent_requests: int = 0
 
 
-SHAREGPT_URL = "https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json"
+SHAREGPT_REPO_ID = "anon8231489123/ShareGPT_Vicuna_unfiltered"
+SHAREGPT_FILENAME = "ShareGPT_V3_unfiltered_cleaned_split.json"
 MOONCAKE_DATASET_URL = {
     "mooncake": "https://raw.githubusercontent.com/kvcache-ai/Mooncake/main/FAST25-release/arxiv-trace/mooncake_trace.jsonl",
     "conversation": "https://raw.githubusercontent.com/kvcache-ai/Mooncake/main/FAST25-release/traces/conversation_trace.jsonl",
@@ -927,6 +928,19 @@ MOONCAKE_DATASET_URL = {
 }
 
 
+def download_and_cache_hf_file(
+    repo_id: str,
+    filename: str,
+    repo_type: str = "dataset",
+):
+    """Download a file from Hugging Face and cache it locally."""
+    from huggingface_hub import hf_hub_download
+
+    return hf_hub_download(
+        repo_id=repo_id, filename=filename, repo_type=repo_type
+    )
+
+
 def download_and_cache_file(url: str, filename: Optional[str] = None):
     """Read and cache a file from a url."""
     if filename is None:
@@ -1171,7 +1185,10 @@ def sample_sharegpt_requests(
 
     # Download sharegpt if necessary
     if not is_file_valid_json(dataset_path) and dataset_path == "":
-        dataset_path = download_and_cache_file(SHAREGPT_URL)
+        dataset_path = download_and_cache_hf_file(
+            repo_id=SHAREGPT_REPO_ID,
+            filename=SHAREGPT_FILENAME,
+        )
 
     # Load the dataset.
     with open(dataset_path) as f:
@@ -1283,7 +1300,10 @@ def sample_random_requests(
 
         # Download sharegpt if necessary
         if not is_file_valid_json(dataset_path):
-            dataset_path = download_and_cache_file(SHAREGPT_URL)
+            dataset_path = download_and_cache_hf_file(
+                repo_id=SHAREGPT_REPO_ID,
+                filename=SHAREGPT_FILENAME,
+            )
 
         # Load the dataset.
         with open(dataset_path) as f: