diff --git a/check_model_size.py b/check_model_size.py
index 76c9cab..f544a6b 100644
--- a/check_model_size.py
+++ b/check_model_size.py
@@ -43,14 +43,11 @@ print(f"BnB 4-bit parameters: {bnb_params / 1e9:.2f}B")
 print(f"BF16 parameters: {bf16_params / 1e9:.2f}B")
 print(f"Estimated size: {(bnb_params * 0.5 + bf16_params * 2) / 1e9:.2f} GB")
 
-# Try to move to GPU
-print("\n   Moving to GPU 0...")
-try:
-    model = model.to("cuda:0")
-    print(f"✓ Success! GPU 0: {torch.cuda.memory_allocated(0) / 1e9:.2f} GB")
-    print(f"   Free VRAM: {torch.cuda.get_device_properties(0).total_memory / 1e9 - torch.cuda.memory_allocated(0) / 1e9:.2f} GB")
-except Exception as e:
-    print(f"✗ FAILED: {e}")
+if bnb_params / total_params > 0.9:
+    print("\n✓ SUCCESS: Model is properly quantized to 4-bit!")
+else:
+    print(f"\n✗ FAILED: Only {bnb_params/total_params*100:.1f}% of parameters are 4-bit")
+    print("   Expected: ~100%, Got: this percentage")
 
 del model
 import gc