diff --git a/lambench/tasks/calculator/calculator_tasks.yml b/lambench/tasks/calculator/calculator_tasks.yml
index 0b62b7b..718e3b3 100644
--- a/lambench/tasks/calculator/calculator_tasks.yml
+++ b/lambench/tasks/calculator/calculator_tasks.yml
@@ -12,6 +12,7 @@ inference_efficiency:
   test_data: /bohr/lambench-efficiency-rg7a/v3/efficiency
   calculator_params:
     warmup_ratio: 0.1
+    natoms_upper_limit: 850
 torsionnet:
   test_data: /bohr/lambench-torsionnet-e4sc/v2/torsionnet500_wB97m
   calculator_params: null
diff --git a/lambench/tasks/calculator/inference_efficiency/inference_efficiency.py b/lambench/tasks/calculator/inference_efficiency/inference_efficiency.py
index ff8871a..3cd52b4 100644
--- a/lambench/tasks/calculator/inference_efficiency/inference_efficiency.py
+++ b/lambench/tasks/calculator/inference_efficiency/inference_efficiency.py
@@ -16,7 +16,7 @@
 
 
 def run_inference(
-    model: ASEModel, test_data: Path, warmup_ratio: float
+    model: ASEModel, test_data: Path, warmup_ratio: float, natoms_upper_limit: int
 ) -> dict[str, dict[str, float]]:
     """
     Inference for all trajectories, return average time and success rate for each system.
@@ -26,7 +26,9 @@ def run_inference(
     for traj in trajs:
         system_name = traj.name
         try:
-            system_result = run_one_inference(model, traj, warmup_ratio)
+            system_result = run_one_inference(
+                model, traj, warmup_ratio, natoms_upper_limit
+            )
             average_time = system_result["average_time"]
             std_time = system_result["std_time"]
             success_rate = system_result["success_rate"]
@@ -52,6 +54,7 @@ def run_one_inference(
     model: ASEModel,
     test_traj: Path,
     warmup_ratio: float,
+    natoms_upper_limit: int,
 ) -> dict[str, float]:
     """
     Infer for one trajectory, return averaged time and success rate, starting timing at warmup_ratio.
@@ -65,7 +68,7 @@ def run_one_inference(
     efficiency = []
     for i, atoms in enumerate(test_atoms):
         # find maximum allowed natoms
-        max_natoms = binary_search_max_natoms(model, atoms)
+        max_natoms = binary_search_max_natoms(model, atoms, natoms_upper_limit)
         # on-the-fly expand atoms
         scaling_factor = np.int32(np.floor(max_natoms / len(atoms)))
         while 1 in find_even_factors(scaling_factor) and scaling_factor > 1: