diff --git a/gpu-validation/defaults/main.yaml b/gpu-validation/defaults/main.yaml
index 59f8ba3..a12b4c4 100644
--- a/gpu-validation/defaults/main.yaml
+++ b/gpu-validation/defaults/main.yaml
@@ -75,3 +75,5 @@ gpu_validation_workload_userns:  # "--userns=keep-id:uid=1001" for RHAIIS
 gpu_validation_workload_shm_size: "--shm-size=4g"
 # [string] Additional environment variables
 gpu_validation_workload_additional_env: "--env=HF_HUB_OFFLINE=0 --env=VLLM_NO_USAGE_STATS=1"
+# [string] vLLM API URL
+gpu_validation_vllm_api_url: "http://127.0.0.1:8000"
diff --git a/gpu-validation/tasks/model_download_and_serve.yaml b/gpu-validation/tasks/model_download_and_serve.yaml
index fa802df..275316e 100644
--- a/gpu-validation/tasks/model_download_and_serve.yaml
+++ b/gpu-validation/tasks/model_download_and_serve.yaml
@@ -57,7 +57,7 @@
 
 - name: Wait for vllm API to appear
   ansible.builtin.uri:
-    url: http://localhost:8000/health
+    url: "{{ gpu_validation_vllm_api_url }}/health"
   register: api_result
   until: ("status" in api_result) and (api_result.status == 200)
   retries: 180  # 180 * 10 = 30 mins - this includes the time to download the model
@@ -66,7 +66,7 @@
 
 - name: Wait for vllm metrics endpoint to appear
   ansible.builtin.uri:
-    url: http://localhost:8000/metrics/
+    url: "{{ gpu_validation_vllm_api_url }}/metrics/"
   register: metrics_result
   until: ("status" in metrics_result) and (metrics_result.status == 200)
   retries: 12  # 12 * 10 = 2 mins after API appears
diff --git a/gpu-validation/tasks/model_performance.yaml b/gpu-validation/tasks/model_performance.yaml
index 5b73208..ee45e60 100644
--- a/gpu-validation/tasks/model_performance.yaml
+++ b/gpu-validation/tasks/model_performance.yaml
@@ -2,6 +2,7 @@
 - name: Run the model performance check script
   ansible.builtin.command: /tmp/scripts/model_performance_check.sh
   environment:
+    URL: "{{ gpu_validation_vllm_api_url }}"
     MODEL_NAME: "{{ gpu_validation_model_name }}"
   register: performance_script_output
   changed_when: false