diff --git a/byte_micro_perf/backends/GPU/backend_gpu.py b/byte_micro_perf/backends/GPU/backend_gpu.py
index 651bb903..f5ce1d67 100644
--- a/byte_micro_perf/backends/GPU/backend_gpu.py
+++ b/byte_micro_perf/backends/GPU/backend_gpu.py
@@ -102,10 +102,6 @@ def alltoall(self):
         self.setup_2d_group()
         self.op = AllToAllOp(self.group)
 
-    def broadcast(self):
-        self.setup_2d_group()
-        self.op = BroadcastOp(self.group)
-
     def host2device(self):
         self.op = Host2DeviceOp(torch.device("cuda"))
 
diff --git a/byte_micro_perf/backends/backend.py b/byte_micro_perf/backends/backend.py
index 595a7310..4491e616 100644
--- a/byte_micro_perf/backends/backend.py
+++ b/byte_micro_perf/backends/backend.py
@@ -113,9 +113,6 @@ def reducescatter(self):
     def alltoall(self):
         pass
 
-    def broadcast(self):
-        pass
-
     def host2device(self):
         pass
 
diff --git a/byte_micro_perf/backends/module_store.py b/byte_micro_perf/backends/module_store.py
index 535ff78e..0f265291 100644
--- a/byte_micro_perf/backends/module_store.py
+++ b/byte_micro_perf/backends/module_store.py
@@ -201,16 +201,6 @@ def forward(self, in_tensors_list, out_tensors_list):
         return True
 
 
-class BroadcastOp(torch.nn.Module):
-    def __init__(self, group):
-        super().__init__()
-        self.group = group
-
-    def forward(self, input_tensors):
-        dist.broadcast(input_tensors, 0, self.group)
-        return True
-    
-
 class Device2HostOp(torch.nn.Module):
     def __init__(self):
         super().__init__()
diff --git a/byte_micro_perf/core/perf_engine.py b/byte_micro_perf/core/perf_engine.py
index ff1155f9..9fe744f0 100644
--- a/byte_micro_perf/core/perf_engine.py
+++ b/byte_micro_perf/core/perf_engine.py
@@ -133,7 +133,7 @@ def start_engine(self) -> None:
         output_dir = os.path.abspath("reports/" + self.backend_type)
         os.makedirs(output_dir, exist_ok=True)
 
-        if self.args.task in ["allreduce", "allgather", "reducescatter", "alltoall", "broadcast"]:
+        if self.args.task in ["allreduce", "allgather", "reducescatter", "alltoall"]:
             for group in self.workload["group"]:
                 mp.spawn(fn=self.init_process, args=(group,), nprocs=group)
         else:
diff --git a/byte_micro_perf/workloads/broadcast.json b/byte_micro_perf/workloads/broadcast.json
deleted file mode 100644
index ea00ce86..00000000
--- a/byte_micro_perf/workloads/broadcast.json
+++ /dev/null
@@ -1,30 +0,0 @@
-{
-  "operator": "broadcast", 
-  "iterations": 100,
-  "input_shape_list": [
-    [
-      1024,
-      1024
-    ],
-    [
-      8,
-      1024,
-      1024
-    ],
-    [
-      64,
-      1024,
-      1024
-    ]
-  ],
-  "dtype": [
-    "float32",
-    "bfloat16",
-    "half"
-  ],
-  "group": [
-    2,
-    4,
-    8
-  ]
-}
\ No newline at end of file