diff --git a/demo/app.py b/demo/app.py
index 7dbc59f..b0918f1 100644
--- a/demo/app.py
+++ b/demo/app.py
@@ -19,7 +19,7 @@ vl_gpt = vl_gpt.to(torch.bfloat16).cuda()
 
 vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
 tokenizer = vl_chat_processor.tokenizer
-cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu'
+cuda_device = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'
 # Multimodal Understanding function
 @torch.inference_mode()
 # Multimodal Understanding function
diff --git a/demo/app_janusflow.py b/demo/app_janusflow.py
index 4777196..01f3b70 100644
--- a/demo/app_janusflow.py
+++ b/demo/app_janusflow.py
@@ -5,7 +5,13 @@ from PIL import Image
 from diffusers.models import AutoencoderKL
 import numpy as np
 
-cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu'
+cuda_device = 'cpu' 
+if torch.cuda.is_available():
+    cuda_device = 'cuda'
+elif torch.backends.mps.is_available():
+    cuda_device = 'mps'
+else:
+    cuda_device = 'cpu'
 
 # Load model and processor
 model_path = "deepseek-ai/JanusFlow-1.3B"
diff --git a/demo/fastapi_app.py b/demo/fastapi_app.py
index c2e5710..a789b61 100644
--- a/demo/fastapi_app.py
+++ b/demo/fastapi_app.py
@@ -21,7 +21,7 @@ vl_gpt = vl_gpt.to(torch.bfloat16).cuda()
 
 vl_chat_processor = VLChatProcessor.from_pretrained(model_path)
 tokenizer = vl_chat_processor.tokenizer
-cuda_device = 'cuda' if torch.cuda.is_available() else 'cpu'
+cuda_device = 'cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu'
 
 
 @torch.inference_mode()