diff --git a/tests/unit/test_d1_hd512_only.py b/tests/unit/test_d1_hd512_only.py
index 8b3a4767..7e3c8cb6 100644
--- a/tests/unit/test_d1_hd512_only.py
+++ b/tests/unit/test_d1_hd512_only.py
@@ -46,11 +46,7 @@ def test():
     import time
     t0 = time.time()
     from cutlass.base_dsl.compiler import PtxasOptions, OptLevel
-    # OptLevel(0): skip MLIR optimizations for faster compilation.
-    # The bottleneck is the MLIR optimizer (not ptxas), so ptxas -j doesn't help.
-    # Verify correctness first at O0, then re-compile at O3 for production.
-    compiled = cute.compile(kernel, mQ, mK, mV, mC, stream, mLSE,
-                           options="--opt-level 0")
+    compiled = cute.compile(kernel, mQ, mK, mV, mC, stream, mLSE)
     t1 = time.time()
     print(f'Compilation took {t1-t0:.1f}s', flush=True)