diff --git a/README.md b/README.md
index 1a62faf..20474b0 100644
--- a/README.md
+++ b/README.md
@@ -261,7 +261,13 @@ boundary statement.
 
 ---
 
-## Performance snapshot *(Ryzen 7950X, Linux 6.9)*
+## Performance snapshot
+
+These are indicative figures from one reference machine *(Ryzen 7950X, Linux
+6.9, sub-interpreter backend)* — hardware-, kernel-, and build-dependent, and
+**not** a benchmark to copy into a comparison. Reproduce them on your own host
+with `python scripts/benchmark.py` (add `--backend process` for the process
+boundary); the encrypted-throughput and RSS rows are not yet covered by it.
 
 | Metric                  | Value   |
 | ----------------------- | ------- |
diff --git a/scripts/benchmark.py b/scripts/benchmark.py
new file mode 100755
index 0000000..a8c0061
--- /dev/null
+++ b/scripts/benchmark.py
@@ -0,0 +1,102 @@
+#!/usr/bin/env python3
+"""Reproducible micro-benchmarks for PyIsolate on the current host.
+
+Reports spawn latency and cell round-trip time for the sub-interpreter backend
+(and, with ``--backend process``, the process backend). The numbers are
+hardware-, kernel-, and build-dependent, so run this on your own machine rather
+than trusting a headline figure copied from someone else's.
+
+Usage::
+
+    python scripts/benchmark.py
+    python scripts/benchmark.py --backend process --iterations 500
+"""
+
+from __future__ import annotations
+
+import argparse
+import statistics
+import sys
+import time
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+import pyisolate as iso
+
+
+def bench_spawn(iterations: int, backend: str) -> list[float]:
+    """Return per-op spawn+close wall times in milliseconds."""
+    samples: list[float] = []
+    for i in range(iterations):
+        start = time.perf_counter()
+        sb = iso.spawn(f"bench-spawn-{i}", backend=backend)
+        sb.close()
+        samples.append((time.perf_counter() - start) * 1e3)
+    return samples
+
+
+def bench_roundtrip(iterations: int, backend: str) -> list[float]:
+    """Return per-op exec+recv round-trip times in microseconds."""
+    samples: list[float] = []
+    with iso.spawn("bench-rt", backend=backend) as sb:
+        # Warm up so the first-call cost does not skew the distribution.
+        sb.exec("post(1)")
+        sb.recv(timeout=5)
+        for _ in range(iterations):
+            start = time.perf_counter()
+            sb.exec("post(1)")
+            sb.recv(timeout=5)
+            samples.append((time.perf_counter() - start) * 1e6)
+    return samples
+
+
+def _summary(samples: list[float]) -> dict[str, float]:
+    ordered = sorted(samples)
+    p95 = ordered[min(len(ordered) - 1, int(len(ordered) * 0.95))]
+    return {
+        "mean": statistics.fmean(samples),
+        "median": statistics.median(samples),
+        "p95": p95,
+        "min": ordered[0],
+        "max": ordered[-1],
+    }
+
+
+def main(argv: list[str] | None = None) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--backend",
+        default="subinterpreter",
+        choices=["subinterpreter", "process"],
+        help="isolation backend to benchmark (default: subinterpreter)",
+    )
+    parser.add_argument(
+        "--iterations",
+        type=int,
+        default=200,
+        help="samples per benchmark (default: 200)",
+    )
+    args = parser.parse_args(argv)
+
+    print(f"PyIsolate benchmark  backend={args.backend}  n={args.iterations}")
+    print(f"python={sys.version.split()[0]}  platform={sys.platform}\n")
+
+    spawn = _summary(bench_spawn(args.iterations, args.backend))
+    rt = _summary(bench_roundtrip(args.iterations, args.backend))
+
+    print(f"{'metric':<22}{'mean':>10}{'median':>10}{'p95':>10}")
+    print(
+        f"{'spawn latency (ms)':<22}"
+        f"{spawn['mean']:>10.3f}{spawn['median']:>10.3f}{spawn['p95']:>10.3f}"
+    )
+    print(
+        f"{'round-trip (us)':<22}"
+        f"{rt['mean']:>10.1f}{rt['median']:>10.1f}{rt['p95']:>10.1f}"
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
diff --git a/tests/test_benchmark.py b/tests/test_benchmark.py
new file mode 100644
index 0000000..bde8a80
--- /dev/null
+++ b/tests/test_benchmark.py
@@ -0,0 +1,39 @@
+"""Smoke tests for scripts/benchmark.py (importability + summary stats).
+
+The spawn/round-trip measurements are covered indirectly by the suite's
+performance tests; here we only lock the script's importability and the pure
+statistics helper so the benchmark does not silently bit-rot.
+"""
+
+import importlib.util
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+
+def _load_benchmark():
+    spec = importlib.util.spec_from_file_location(
+        "pyisolate_benchmark", ROOT / "scripts" / "benchmark.py"
+    )
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_summary_reports_expected_statistics():
+    bench = _load_benchmark()
+    summary = bench._summary([10.0, 20.0, 30.0, 40.0])
+    assert summary["min"] == 10.0
+    assert summary["max"] == 40.0
+    assert summary["mean"] == 25.0
+    assert summary["median"] == 25.0
+    assert summary["p95"] >= summary["median"]
+
+
+def test_benchmark_exposes_expected_entry_points():
+    bench = _load_benchmark()
+    assert callable(bench.bench_spawn)
+    assert callable(bench.bench_roundtrip)
+    assert callable(bench.main)