diff --git a/.github/workflows/persistence-performance.yml b/.github/workflows/persistence-performance.yml
new file mode 100644
index 00000000..6cc6aa7c
--- /dev/null
+++ b/.github/workflows/persistence-performance.yml
@@ -0,0 +1,26 @@
+name: HF integration tests
+
+on:
+  schedule:
+    - cron:  '0 9 * * 0'  # every sunday at 9:00 UTC
+  workflow_dispatch:
+
+jobs:
+  check-persistence-performance:
+
+    runs-on: ubuntu-latest
+    if: "github.repository == 'skops-dev/skops'"
+
+    timeout-minutes: 10
+
+    steps:
+    - uses: actions/checkout@v3
+    - name: Set up Python
+      uses: actions/setup-python@v4
+    - name: Install requirements
+      run: |
+        pip install .[tests]
+        pip --version
+        pip list
+    - name: Run persistence performance checks
+      run: python scripts/check_persistence_performance.py
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 0bca976d..d95e8edf 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -1,6 +1,6 @@
 repos:
 -   repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.1.0
+    rev: v4.4.0
     hooks:
     -   id: check-yaml
         exclude: .github/conda/meta.yaml
@@ -10,11 +10,11 @@ repos:
     -   id: check-case-conflict
     -   id: check-merge-conflict
 -   repo: https://github.com/psf/black
-    rev: 22.6.0
+    rev: 23.1.0
     hooks:
     -   id: black
 -   repo: https://github.com/pycqa/flake8
-    rev: 4.0.1
+    rev: 6.0.0
     hooks:
     -   id: flake8
         types: [file, python]
@@ -23,7 +23,7 @@ repos:
     hooks:
     -   id: isort
 -   repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.0.0
+    rev: v1.0.1
     hooks:
     -   id: mypy
         args: [--config-file=pyproject.toml]
diff --git a/scripts/check_persistence_performance.py b/scripts/check_persistence_performance.py
new file mode 100644
index 00000000..5f6d2ca1
--- /dev/null
+++ b/scripts/check_persistence_performance.py
@@ -0,0 +1,117 @@
+"""Check that the performance of skops persistence is not too slow
+
+Load each (fitted) estimator and persist it with pickle and with skops. Measure
+the time it takes and record it. Report the estimators that were slowest. If
+skops is much slower than pickle (in absolute terms), raise an error to make the
+GH action fail.
+
+"""
+
+from __future__ import annotations
+
+import pickle
+import timeit
+import warnings
+from typing import Any
+
+import pandas as pd
+from sklearn.utils._tags import _safe_tags
+from sklearn.utils._testing import set_random_state
+
+import skops.io as sio
+from skops.io.tests.test_persist import (
+    _get_check_estimator_ids,
+    _tested_estimators,
+    get_input,
+)
+
+ATOL = 1  # seconds absolute difference allowed at max
+NUM_REPS = 10  # number of times the check is repeated
+TOPK = 10  # number of slowest estimators reported
+
+
+def check_persist_performance() -> None:
+    """Run all performance checks on all estimators and print results.
+
+    For each estimator, record how long it takes to dump+load with pickle and
+    with skops. If any estimator takes much longer (in absolute time) with skops
+    than pickle, raise a RuntimeError. Print the worst results to sdtout.
+
+    """
+    results: dict[str, list[Any]] = {"name": [], "pickle (s)": [], "skops (s)": []}
+    for estimator in _tested_estimators():
+        set_random_state(estimator, random_state=0)
+
+        X, y = get_input(estimator)
+        tags = _safe_tags(estimator)
+        if tags.get("requires_fit", True):
+            with warnings.catch_warnings():
+                warnings.filterwarnings("ignore", module="sklearn")
+                if y is not None:
+                    estimator.fit(X, y)
+                else:
+                    estimator.fit(X)
+
+        name = _get_check_estimator_ids(estimator)
+        time_pickle, time_skops = run_check(estimator, number=NUM_REPS)
+
+        results["name"].append(name)
+        results["pickle (s)"].append(time_pickle)
+        results["skops (s)"].append(time_skops)
+
+    format_result(results, topk=TOPK)
+
+
+def run_check(estimator, number: int) -> tuple[float, float]:
+    """Run performance check with the given estimator for pickle and skops.
+
+    The test is run multiple times to get more robust results, ``number``
+    indicates how often the it is run.
+
+    """
+
+    def run_pickle():
+        pickle.loads(pickle.dumps(estimator))
+
+    def run_skops():
+        sio.loads(sio.dumps(estimator), trusted=True)
+
+    time_pickle = timeit.timeit(run_pickle, number=number) / number
+    time_skops = timeit.timeit(run_skops, number=number) / number
+
+    return time_pickle, time_skops
+
+
+def format_result(results: dict[str, list[Any]], topk: int) -> None:
+    """Report results from performance checks.
+
+    Print the ``topk`` slowest results. If any estimator takes much longer (in
+    absolute time) with skops than pickle, raise a RuntimeError.
+
+    """
+    df = pd.DataFrame(results)
+    df = df.assign(
+        abs_diff=df["skops (s)"] - df["pickle (s)"],
+        rel_diff=df["skops (s)"] / df["pickle (s)"],
+        too_slow=lambda d: d["abs_diff"] > ATOL,
+    )
+
+    df = df.sort_values(["abs_diff"], ascending=False).reset_index(drop=True)
+    print(f"{topk} largest differences:")
+    print(df.head(10))
+
+    df_slow = df.query("too_slow")
+    if df_slow.empty:
+        print("No estimator was found to be unacceptably slow")
+        return
+
+    print(
+        f"Found {len(df_slow)} estimator(s) that are at least {ATOL:.1f} sec slower "
+        "with skops:"
+    )
+    print(", ".join(df_slow["name"].tolist()))
+    raise RuntimeError("Skops persistence too slow")
+
+
+if __name__ == "__main__":
+    check_persist_performance()