diff --git a/benchmarks/backend_request_func.py b/benchmarks/backend_request_func.py
index 837d9df91e9..596804331f2 100644
--- a/benchmarks/backend_request_func.py
+++ b/benchmarks/backend_request_func.py
@@ -128,13 +128,13 @@ async def async_request_eb_openai_chat_completions(
 
                         chunk = chunk_bytes.decode("utf-8").removeprefix("data: ")
                         if chunk != "[DONE]":
-                            # print("####chunk:", chunk, type(chunk))
+                            #print("####chunk:", chunk, type(chunk))
                             timestamp = time.perf_counter()
                             data = json.loads(chunk)
 
                             if request_id == "None" and "id" in data:
                                 request_id = data["id"]
-
+                            
                             if choices := data.get("choices"):
                                 content = choices[0]["delta"].get("content")
                                 reason_content = choices[0]["delta"].get("reasoning_content")
@@ -143,9 +143,12 @@ async def async_request_eb_openai_chat_completions(
                                     ttft = timestamp - st
                                     output.ttft = ttft
                                     # cached_tokens
-                                    output.prompt_len = (
-                                        data["usage"].get("prompt_tokens_details", {}).get("cached_tokens", 0)
-                                    )
+                                    if data["usage"] and data["usage"].get("prompt_tokens_details", {}):
+                                        output.prompt_len = (
+                                            data["usage"].get("prompt_tokens_details", {}).get("cached_tokens", 0)
+                                        )
+                                    else:
+                                        output.prompt_len = 0
 
                                 # Decoding phase
                                 else:
@@ -157,6 +160,7 @@ async def async_request_eb_openai_chat_completions(
                             elif usage := data.get("usage", {}):
                                 output.output_tokens = usage.get("completion_tokens", 0)
                                 output.prompt_tokens = usage.get("prompt_tokens", 0)
+                            
 
                             most_recent_timestamp = timestamp
 
diff --git a/benchmarks/benchmark_serving.py b/benchmarks/benchmark_serving.py
index ce072555fa7..fb13301c4e4 100644
--- a/benchmarks/benchmark_serving.py
+++ b/benchmarks/benchmark_serving.py
@@ -635,7 +635,7 @@ def benchmark_metrics(
     goodput_config_dict = check_goodput_args(args)
 
     metrics, actual_output_lens = calculate_metrics(
-        input_requests=input_requests,
+        # input_requests=input_requests,
         outputs=outputs,
         dur_s=benchmark_duration,
         selected_percentiles=selected_percentiles,
diff --git a/benchmarks/yaml/request_yaml/vLLM_default.yaml b/benchmarks/yaml/request_yaml/request.yaml
similarity index 53%
rename from benchmarks/yaml/request_yaml/vLLM_default.yaml
rename to benchmarks/yaml/request_yaml/request.yaml
index a6385823b5f..9fc603354b6 100644
--- a/benchmarks/yaml/request_yaml/vLLM_default.yaml
+++ b/benchmarks/yaml/request_yaml/request.yaml
@@ -1,11 +1,11 @@
-top_p: 1.0
-temperature: 1.0
-metadata:
-  min_tokens: 1
-max_tokens: 30721
+top_p: 0.8
+temperature: 0.8
+max_tokens: 12288
 repetition_penalty: 1.0
 frequency_penalty: 0
 presence_penalty: 0
-skip_special_tokens: false
+metadata:
+  enable_thinking: false
+  min_tokens: 1
 chat_template_kwargs:
-  enable_thinking: true
+  enable_thinking: false