diff --git a/.github/workflows/benchmark.yml b/.github/workflows/benchmark.yml
new file mode 100644
index 0000000..1c9f40c
--- /dev/null
+++ b/.github/workflows/benchmark.yml
@@ -0,0 +1,65 @@
+name: Run benchmark
+
+on:
+  pull_request:
+    branches: [master, develop]
+
+permissions:
+  contents: read
+  pull-requests: write
+
+jobs:
+  benchmark:
+    name: Performance regression check
+    runs-on: ubuntu-latest
+    steps:
+    - name: Checkout the repo
+      uses: actions/checkout@v4
+      with:
+        # Get history and tags for SCM versioning to work
+        fetch-depth: 0
+    - name: Install the latest version of uv with cache enabled
+      uses: astral-sh/setup-uv@v3
+      with:
+        version: "latest"
+        enable-cache: true
+        cache-dependency-glob: ""
+    - name: Run benchmark on develop and current ref, then compare
+      shell: bash
+      run: |
+        set -euo pipefail
+
+        CURRENT_SHA="$(git rev-parse HEAD)"
+
+        echo "Running benchmark on develop"
+        git checkout --detach origin/develop
+        uvx --python 3.13 --with tox-uv tox -e benchmark -- --benchmark-json baseline.json
+
+        echo "Running benchmark on current ref"
+        git checkout --detach "${CURRENT_SHA}"
+        uvx --python 3.13 --with tox-uv tox -e benchmark -- --benchmark-json candidate.json
+
+        echo "Comparing benchmark results"
+        uvx --python 3.13 \
+          --with "pytest-benchmark @ git+https://github.com/ionelmc/pytest-benchmark.git@master" \
+          pytest-benchmark compare --between=median \
+          baseline.json candidate.json \
+          | tee benchmark-compare.txt
+
+        {
+          echo "## Benchmark comparison"
+          echo
+          echo "Compared baseline: \`origin/develop\`"
+          echo "Compared candidate: ${CURRENT_SHA}"
+          echo
+          echo '```text'
+          cat benchmark-compare.txt
+          echo '```'
+        } > benchmark-comment.md
+
+    - name: Publish benchmark comparison comment
+      if: github.event.pull_request.head.repo.full_name == github.repository
+      uses: marocchino/sticky-pull-request-comment@v2
+      with:
+        header: benchmark-comparison
+        path: benchmark-comment.md
diff --git a/tests/test_benchmark.py b/tests/test_benchmark.py
new file mode 100644
index 0000000..b68a870
--- /dev/null
+++ b/tests/test_benchmark.py
@@ -0,0 +1,144 @@
+# -*- coding: utf-8 -*-
+# Copyright 2026, CS GROUP - France, https://www.csgroup.eu/
+#
+# This file is part of EODAG project
+#     https://www.github.com/CS-SI/EODAG
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import subprocess
+import sys
+
+from tests.context import (
+    TEST_RESOURCES_PATH,
+    AwsDownload,
+    EOProduct,
+    PluginConfig,
+    path_to_uri,
+)
+from tests.integration import test_eoproduct_xarray
+
+
+def _run_unittest_method(test_case_cls, method_name):
+    """Run one unittest test method with isolated setup/teardown."""
+    test_case = test_case_cls(methodName=method_name)
+    test_case.setUp()
+    try:
+        getattr(test_case, method_name)()
+    finally:
+        test_case.tearDown()
+
+
+def _import_eodag_cube_subprocess():
+    """Import eodag_cube in a subprocess (cold start / import cost path)."""
+    result = subprocess.run(
+        [sys.executable, "-c", "import eodag_cube"],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    assert result.returncode == 0, result.stderr
+
+
+def _import_eodag_public_api_subprocess():
+    """Resolve eodag's public product classes in a subprocess.
+
+    When eodag-cube is installed, plain eodag silently resolves
+    ``eodag.api.product.EOProduct``/``Asset``/``AssetsDict`` to the eodag-cube
+    subclasses (see ``eodag.api.product._resolve``). This measures the
+    import/resolution cost that plain eodag users pay just by having eodag-cube
+    installed, without calling any eodag-cube-specific method.
+    """
+    result = subprocess.run(
+        [
+            sys.executable,
+            "-c",
+            (
+                "import eodag;"
+                "from eodag.api.product import EOProduct, Asset, AssetsDict;"
+                "assert EOProduct.__module__.startswith('eodag_cube'), EOProduct.__module__"
+            ),
+        ],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    assert result.returncode == 0, result.stderr
+
+
+def test_benchmark_import_eodag_cube_subprocess(benchmark):
+    benchmark.pedantic(
+        _import_eodag_cube_subprocess,
+        rounds=10,
+        iterations=1,
+    )
+
+
+def test_benchmark_import_eodag_public_api_subprocess(benchmark):
+    benchmark.pedantic(
+        _import_eodag_public_api_subprocess,
+        rounds=10,
+        iterations=1,
+    )
+
+
+def test_benchmark_eoproduct_instantiation(benchmark):
+    # ``EOProduct`` here is the class plain eodag also resolves to when eodag-cube
+    # is installed. eodag-cube's ``__init__`` rebuilds the assets through its own
+    # ``AssetsDict``, so this captures the per-product side-effect paid even when
+    # no eodag-cube-specific method (to_xarray/get_data) is ever called.
+    case = test_eoproduct_xarray.TestEOProductXarray(methodName="test_to_xarray_local")
+    case.setUp()
+    try:
+        benchmark(lambda: EOProduct(case.provider, case.eoproduct_props, collection=case.collection))
+    finally:
+        case.tearDown()
+
+
+def test_benchmark_eoproduct_assets_population(benchmark):
+    # Populating assets goes through eodag-cube's ``AssetsDict.__setitem__``, which
+    # wraps every entry in eodag-cube's ``Asset``. This is a side-effect on plain
+    # eodag asset construction (e.g. when parsing search results), with no xarray
+    # or rasterio usage involved.
+    case = test_eoproduct_xarray.TestEOProductXarray(methodName="test_to_xarray_local")
+    case.setUp()
+    assets = {f"band{i}": {"href": f"http://example.com/band{i}.tif"} for i in range(20)}
+
+    def _populate_assets():
+        product = EOProduct(case.provider, case.eoproduct_props, collection=case.collection)
+        product.assets.update(assets)
+        return len(product.assets)
+
+    try:
+        benchmark(_populate_assets)
+    finally:
+        case.tearDown()
+
+
+def test_benchmark_to_xarray_local(benchmark):
+    case = test_eoproduct_xarray.TestEOProductXarray(methodName="test_to_xarray_local")
+    case.setUp()
+    products_path = os.path.join(TEST_RESOURCES_PATH, "products")
+
+    def _to_xarray_local():
+        product = EOProduct(case.provider, case.eoproduct_props, collection=case.collection)
+        product.register_downloader(AwsDownload("foo", PluginConfig()), None)
+        product.location = path_to_uri(products_path)
+        with product.to_xarray() as xarray_dict:
+            return len(xarray_dict)
+
+    try:
+        benchmark(_to_xarray_local)
+    finally:
+        case.tearDown()
diff --git a/tox.ini b/tox.ini
index 5f2ecf4..a9c63c4 100644
--- a/tox.ini
+++ b/tox.ini
@@ -16,7 +16,16 @@ install_command = uv pip install --upgrade {opts} {packages}
 deps =
     .[dev]
 commands =
-    pytest --show-capture=no --cov --cov-fail-under=50 --cov-report=term-missing {posargs}
+    pytest --show-capture=no --cov --cov-fail-under=50 \
+        --cov-report=term-missing {posargs: --ignore=tests/test_benchmark.py}
+
+[testenv:benchmark]
+basepython = python3
+deps =
+    .[dev]
+    pytest-benchmark[histogram] @ git+https://github.com/ionelmc/pytest-benchmark.git@master
+commands =
+    pytest -v tests/test_benchmark.py --benchmark-disable-gc {posargs}
 
 [testenv:pypi]
 allowlist_externals = /bin/bash