diff --git a/timesfm-forecasting/examples/covariates-forecasting/demo_covariates.py b/timesfm-forecasting/examples/covariates-forecasting/demo_covariates.py
index 65153c01..20b9394d 100644
--- a/timesfm-forecasting/examples/covariates-forecasting/demo_covariates.py
+++ b/timesfm-forecasting/examples/covariates-forecasting/demo_covariates.py
@@ -411,9 +411,18 @@ def demonstrate_api() -> None:
 pip install timesfm[xreg]
 
 import timesfm
-hparams   = timesfm.TimesFmHparams(backend="cpu", per_core_batch_size=32, horizon_len=12)
-ckpt      = timesfm.TimesFmCheckpoint(huggingface_repo_id="google/timesfm-2.5-200m-pytorch")
-model     = timesfm.TimesFm(hparams=hparams, checkpoint=ckpt)
+
+model = timesfm.TimesFM_2p5_200M_torch.from_pretrained(
+    "google/timesfm-2.5-200m-pytorch",
+    torch_compile=False,
+)
+model.compile(timesfm.ForecastConfig(
+    max_context=512,
+    max_horizon=12,
+    normalize_inputs=True,
+    use_continuous_quantile_head=True,
+    fix_quantile_crossing=True,
+))
 
 point_fc, quant_fc = model.forecast_with_covariates(
     inputs=[sales_a, sales_b, sales_c],
@@ -423,8 +432,8 @@ def demonstrate_api() -> None:
     xreg_mode="xreg + timesfm",
     normalize_xreg_target_per_input=True,
 )
-# point_fc:  (num_series, horizon_len)
-# quant_fc:  (num_series, horizon_len, 10)
+# point_fc:  (num_series, horizon)
+# quant_fc:  (num_series, horizon, 10)  # columns: [mean, q10, q20, ..., q90]
 """)