diff --git a/src/diffusers/schedulers/scheduling_flow_match_euler_discrete.py b/src/diffusers/schedulers/scheduling_flow_match_euler_discrete.py
index 0e4f5c6a1f97..cdfef1bcbd2e 100644
--- a/src/diffusers/schedulers/scheduling_flow_match_euler_discrete.py
+++ b/src/diffusers/schedulers/scheduling_flow_match_euler_discrete.py
@@ -128,6 +128,13 @@ def __init__(
         timesteps = torch.from_numpy(timesteps).to(dtype=torch.float32)
 
         sigmas = timesteps / num_train_timesteps
+        # Store sigma_min / sigma_max from the *unshifted* sigmas. `set_timesteps` uses these
+        # to seed the per-call schedule and then re-applies `shift` itself — storing the
+        # post-shift values here would make `set_timesteps` shift a second time and produce
+        # sigmas that don't match `__init__`'s schedule (#13243).
+        self.sigma_min = sigmas[-1].item()
+        self.sigma_max = sigmas[0].item()
+
         if not use_dynamic_shifting:
             # when use_dynamic_shifting is True, we apply the timestep shifting on the fly based on the image resolution
             sigmas = shift * sigmas / (1 + (shift - 1) * sigmas)
@@ -140,8 +147,6 @@ def __init__(
         self._shift = shift
 
         self.sigmas = sigmas.to("cpu")  # to avoid too much CPU/GPU communication
-        self.sigma_min = self.sigmas[-1].item()
-        self.sigma_max = self.sigmas[0].item()
 
     @property
     def shift(self):