diff --git a/distribution/build_docs.py b/distribution/build_docs.py
index eee815e4103..7602f347b28 100644
--- a/distribution/build_docs.py
+++ b/distribution/build_docs.py
@@ -4,6 +4,7 @@
 import shutil
 import sys
 import textwrap
+from datetime import datetime
 from os import PathLike, environ
 from pathlib import Path
 from pprint import pprint
@@ -14,12 +15,15 @@
 
 import pytest
 from benchmark import run_benchmarks
+from flaky import flaky
 from modflow_devtools.build import meson_build
 from modflow_devtools.download import (
     download_and_unzip,
+    download_artifact,
     get_release,
+    list_artifacts,
 )
-from modflow_devtools.markers import no_parallel, requires_exe
+from modflow_devtools.markers import no_parallel, requires_exe, requires_github
 from modflow_devtools.misc import run_cmd, run_py_script, set_dir
 
 from utils import assert_match, convert_line_endings, get_project_root_path, glob, match
@@ -67,20 +71,73 @@
 ]
 
 
+def download_benchmarks(
+    output_path: PathLike,
+    verbose: bool = False,
+    repo_owner: str = "MODFLOW-USGS",
+) -> Optional[Path]:
+    """Try to download MF6 benchmarks from GitHub Actions."""
+
+    output_path = Path(output_path).expanduser().absolute()
+    name = "run-time-comparison"  # todo make configurable
+    repo = f"{repo_owner}/modflow6"  # todo make configurable, add pytest/cli args
+    artifacts = list_artifacts(repo, name=name, verbose=verbose)
+    artifacts = sorted(
+        artifacts,
+        key=lambda a: datetime.strptime(a["created_at"], "%Y-%m-%dT%H:%M:%SZ"),
+        reverse=True,
+    )
+    artifacts = [
+        a
+        for a in artifacts
+        if a["workflow_run"]["head_branch"] == "develop"  # todo make configurable
+    ]
+    most_recent = next(iter(artifacts), None)
+    print(f"Found most recent benchmarks (artifact {most_recent['id']})")
+    if most_recent:
+        print(f"Downloading benchmarks (artifact {most_recent['id']})")
+        download_artifact(repo, id=most_recent["id"], path=output_path, verbose=verbose)
+        print(f"Downloaded benchmarks to {output_path}")
+        path = output_path / f"{name}.md"
+        assert path.is_file()
+        return path
+    else:
+        print("No benchmarks found")
+        return None
+
+
 @pytest.fixture
 def github_user() -> Optional[str]:
     return environ.get("GITHUB_USER", None)
 
 
+@flaky
+@no_parallel
+@requires_github
+def test_download_benchmarks(tmp_path, github_user):
+    path = download_benchmarks(
+        tmp_path,
+        verbose=True,
+        repo_owner=github_user if github_user else "MODFLOW-USGS",
+    )
+    if path:
+        assert path.name == "run-time-comparison.md"
+
+
 def build_benchmark_tex(
     output_path: PathLike,
     force: bool = False,
+    repo_owner: str = "MODFLOW-USGS",
 ):
     """Build LaTeX files for MF6 performance benchmarks to go into the release notes."""
 
     BENCHMARKS_PATH.mkdir(parents=True, exist_ok=True)
     benchmarks_path = BENCHMARKS_PATH / "run-time-comparison.md"
 
+    # download benchmark artifacts if any exist on GitHub
+    if not benchmarks_path.is_file():
+        benchmarks_path = download_benchmarks(BENCHMARKS_PATH, repo_owner=repo_owner)
+
     # run benchmarks again if no benchmarks found on GitHub or overwrite requested
     if force or not benchmarks_path.is_file():
         run_benchmarks(
@@ -105,6 +162,20 @@ def build_benchmark_tex(
         assert (RELEASE_NOTES_PATH / f"{benchmarks_path.stem}.tex").is_file()
 
 
+@flaky
+@no_parallel
+@requires_github
+def test_build_benchmark_tex(tmp_path):
+    benchmarks_path = BENCHMARKS_PATH / "run-time-comparison.md"
+    tex_path = DISTRIBUTION_PATH / f"{benchmarks_path.stem}.tex"
+
+    try:
+        build_benchmark_tex(tmp_path)
+        assert benchmarks_path.is_file()
+    finally:
+        tex_path.unlink(missing_ok=True)
+
+
 def build_deprecations_tex(force: bool = False):
     """Build LaTeX files for the deprecations table to go into the release notes."""