diff --git a/runs/test_build_cache.npy b/runs/test_build_cache.npy
new file mode 100644
index 00000000..ef9b3f35
Binary files /dev/null and b/runs/test_build_cache.npy differ
diff --git a/tests/conftest.py b/tests/conftest.py
index 0640d373..f69c9d59 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -1,11 +1,15 @@
 import pytest
+import torch
 from datasets import Dataset
 from transformers import AutoConfig, AutoModelForCausalLM
 
 
 @pytest.fixture
 def model():
-    """Create a small test model."""
+    """Randomly initialize a small test model."""
+    torch.manual_seed(42)
+    torch.cuda.manual_seed(42)
+
     config = AutoConfig.from_pretrained("trl-internal-testing/tiny-Phi3ForCausalLM")
     return AutoModelForCausalLM.from_config(config)
 
diff --git a/tests/test_build.py b/tests/test_build.py
index 3ca98dce..d1cfe16f 100644
--- a/tests/test_build.py
+++ b/tests/test_build.py
@@ -1,22 +1,8 @@
-import pytest
-
-from bergson.data import load_gradients
-
-try:
-    import torch
-
-    HAS_CUDA = torch.cuda.is_available()
-except Exception:
-    HAS_CUDA = False
-
-if not HAS_CUDA:
-    pytest.skip(
-        "Skipping GPU-only tests: no CUDA/NVIDIA driver available.",
-        allow_module_level=True,
-    )
-
 from pathlib import Path
 
+import numpy as np
+import pytest
+import torch
 from transformers import AutoModelForCausalLM
 
 from bergson import (
@@ -24,6 +10,29 @@
     GradientProcessor,
     collect_gradients,
 )
+from bergson.data import load_gradients
+
+
+@pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA not available")
+def test_build_consistency(tmp_path: Path, model, dataset):
+    collect_gradients(
+        model=model,
+        data=dataset,
+        processor=GradientProcessor(),
+        path=str(tmp_path),
+        skip_preconditioners=True,
+    )
+    index = load_gradients(str(tmp_path))
+
+    # Regenerate cache
+    cache_path = Path("runs/test_build_cache.npy")
+    if not cache_path.exists():
+        np.save(cache_path, index[index.dtype.names[0]][0])
+    cached_item_grad = np.load(cache_path)
+
+    first_module_grad = index[index.dtype.names[0]][0]
+
+    assert np.allclose(first_module_grad, cached_item_grad, atol=1e-6)
 
 
 @pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA not available")