diff --git a/gptq/gptj.py b/gptq/gptj.py
index 5ebf7b3e..23f4f96f 100644
--- a/gptq/gptj.py
+++ b/gptq/gptj.py
@@ -535,6 +535,16 @@ def sync():
         quantizers = gptj_sequential(model, dataloader, DEV)
         print(time.time() - tick)
 
+    if args.benchmark:
+        gpus = [torch.device("cuda:%d" % i) for i in range(torch.cuda.device_count())]
+        if len(gpus) > 1:
+            gptj_multigpu(model, gpus)
+        else:
+            model = model.to(DEV)
+        if args.benchmark:
+            input_ids = next(iter(dataloader))[0][:, : args.benchmark]
+            benchmark(model, input_ids, check=args.check)
+
     if args.eval:
         datasets = ["wikitext2", "ptb", "c4"]
         if args.new_eval:
@@ -550,6 +560,9 @@ def sync():
             print(dataset)
             gptj_eval(model, testloader, DEV)
 
+    if args.load:
+        exit()
+
     if args.save:
         gptj_pack(model, quantizers, args.wbits, args.groupsize)
         torch.save(model.state_dict(), args.save)
@@ -559,13 +572,3 @@ def sync():
         from safetensors.torch import save_file as safe_save
 
         safe_save(model.state_dict(), args.save_safetensors)
-
-    if args.benchmark:
-        gpus = [torch.device("cuda:%d" % i) for i in range(torch.cuda.device_count())]
-        if len(gpus) > 1:
-            gptj_multigpu(model, gpus)
-        else:
-            model = model.to(DEV)
-        if args.benchmark:
-            input_ids = next(iter(dataloader))[0][:, : args.benchmark]
-            benchmark(model, input_ids, check=args.check)
diff --git a/gptq/gptneox.py b/gptq/gptneox.py
index 69192b19..ada4c731 100644
--- a/gptq/gptneox.py
+++ b/gptq/gptneox.py
@@ -545,9 +545,6 @@ def sync():
             input_ids = next(iter(dataloader))[0][:, : args.benchmark]
             benchmark(model, input_ids, check=args.check)
 
-    if args.load:
-        exit()
-
     if args.eval:
         datasets = ["wikitext2", "ptb", "c4"]
         if args.new_eval:
@@ -559,6 +556,9 @@ def sync():
             print(dataset)
             gptneox_eval(model, testloader, DEV)
 
+    if args.load:
+        exit()
+
     if args.save:
         gptneox_pack(model, quantizers, args.wbits, args.groupsize)
         torch.save(model.state_dict(), args.save)
diff --git a/gptq/llama.py b/gptq/llama.py
index 30a7dc50..a2c64fd5 100644
--- a/gptq/llama.py
+++ b/gptq/llama.py
@@ -481,9 +481,6 @@ def sync():
         if args.benchmark:
             input_ids = next(iter(dataloader))[0][:, :args.benchmark]
             benchmark(model, input_ids, check=args.check)
-
-    if args.load:
-        exit()
     
     if args.eval:
         datasets = ['wikitext2', 'ptb', 'c4'] 
@@ -496,6 +493,9 @@ def sync():
             print(dataset)
             llama_eval(model, testloader, DEV)
 
+    if args.load:
+        exit()
+
     if args.save:
         llama_pack(model, quantizers, args.wbits, args.groupsize)
         torch.save(model.state_dict(), args.save) 
diff --git a/gptq/mpt.py b/gptq/mpt.py
index b67514df..ec7eae2a 100644
--- a/gptq/mpt.py
+++ b/gptq/mpt.py
@@ -519,6 +519,16 @@ def sync():
         quantizers = mpt_sequential(model, dataloader, DEV)
         print(time.time() - tick)
 
+    if args.benchmark:
+        gpus = [torch.device("cuda:%d" % i) for i in range(torch.cuda.device_count())]
+        if len(gpus) > 1:
+            mpt_multigpu(model, gpus)
+        else:
+            model = model.to(DEV)
+        if args.benchmark:
+            input_ids = next(iter(dataloader))[0][:, : args.benchmark]
+            benchmark(model, input_ids, check=args.check)
+
     if args.eval:
         datasets = ["wikitext2", "ptb", "c4"]
         if args.new_eval:
@@ -530,6 +540,9 @@ def sync():
             print(dataset)
             mpt_eval(model, testloader, DEV)
 
+    if args.load:
+        exit()
+
     if args.save:
         mpt_pack(model, quantizers, args.wbits, args.groupsize)
         torch.save(model.state_dict(), args.save)
@@ -539,15 +552,3 @@ def sync():
         from safetensors.torch import save_file as safe_save
 
         safe_save(model.state_dict(), args.save_safetensors)
-
-    if args.benchmark:
-        gpus = [torch.device("cuda:%d" % i) for i in range(torch.cuda.device_count())]
-        if len(gpus) > 1:
-            mpt_multigpu(model, gpus)
-        else:
-            model = model.to(DEV)
-        if args.benchmark:
-            input_ids = next(iter(dataloader))[0][:, : args.benchmark]
-            benchmark(model, input_ids, check=args.check)
-    if args.load:
-        exit()
diff --git a/gptq/opt.py b/gptq/opt.py
index 5751ae7d..ae9948e5 100644
--- a/gptq/opt.py
+++ b/gptq/opt.py
@@ -484,9 +484,6 @@ def sync():
             input_ids = next(iter(dataloader))[0][:, :args.benchmark]
             benchmark(model, input_ids, check=args.check)
             
-    if args.load:
-        exit()
-
     if args.eval:
         datasets = ['wikitext2', 'ptb', 'c4'] 
         if args.new_eval:
@@ -498,6 +495,9 @@ def sync():
             print(dataset)
             opt_eval(model, testloader, DEV)
 
+    if args.load:
+        exit()
+
     if args.save:
         opt_pack(model, quantizers, args.wbits, args.groupsize)
         torch.save(model.state_dict(), args.save)