diff --git a/onnxscript/function_libs/torch_lib/ops/fft.py b/onnxscript/function_libs/torch_lib/ops/fft.py
index 30d086a5d5..d426d3cd50 100644
--- a/onnxscript/function_libs/torch_lib/ops/fft.py
+++ b/onnxscript/function_libs/torch_lib/ops/fft.py
@@ -172,17 +172,33 @@ def aten__fft_r2c(
         transformed = op.Unsqueeze(self, axes=[-1])
 
     for idx, dimension in enumerate(reversed(dim)):
+        dft_length = op.Squeeze(
+            op.Shape(transformed, start=dimension, end=dimension + 1),
+            axes=[0],
+        )
         transformed = _fftn_onnx_normalization(
             transformed,
             normalization,
-            op.Shape(transformed, start=dimension, end=dimension + 1),
+            dft_length,
             inverse=False,
         )
         if idx > 0:
-            transformed = op.DFT(transformed, axis=dimension, inverse=False, onesided=False)
+            transformed = op.DFT(
+                transformed,
+                dft_length=dft_length,
+                axis=dimension,
+                inverse=False,
+                onesided=False,
+            )
         else:
             # Torch computes one-sided FFT on the last dimension only.
-            transformed = op.DFT(transformed, axis=dimension, inverse=False, onesided=onesided)
+            transformed = op.DFT(
+                transformed,
+                dft_length=dft_length,
+                axis=dimension,
+                inverse=False,
+                onesided=onesided,
+            )
 
     if unsqueeze_first_dim:
         transformed = op.Squeeze(transformed, axes=[0])
diff --git a/tests/function_libs/torch_lib/e2e_ops_tests.py b/tests/function_libs/torch_lib/e2e_ops_tests.py
index b07f57ce20..e5b040b001 100644
--- a/tests/function_libs/torch_lib/e2e_ops_tests.py
+++ b/tests/function_libs/torch_lib/e2e_ops_tests.py
@@ -387,6 +387,23 @@ def forward(self, x):
         )
         _testing.assert_onnx_program(onnx_program)
 
+    def test_rfft_produces_correct_dft_length(self):
+        class RFFTModel(torch.nn.Module):
+            def forward(self, x):
+                x = torch.fft.rfft(x, n=512, dim=-1)
+                return x.real**2 + x.imag**2
+
+        x = torch.randn(4, 512, dtype=torch.float32)
+        onnx_program = torch.onnx.export(
+            RFFTModel(),
+            (x,),
+            opset_version=20,
+            dynamo=True,
+            optimize=False,
+        )
+
+        self.assertEqual(onnx_program.model.graph.outputs[0].shape, [4, 512 // 2 + 1])
+
     def test_avg_pool(self):
         class Model(torch.nn.Module):
             def forward(self, x2d, x3d, x4d, x5d):