diff --git a/.coveragerc b/.coveragerc
index 3f92bc0..b3916b3 100644
--- a/.coveragerc
+++ b/.coveragerc
@@ -30,3 +30,4 @@ source = ijson
 [report]
 exclude_lines =
     raise ImportError\('no backends available'\)
+	 if stdout.isatty():
diff --git a/test/test_benchmark.py b/test/test_benchmark.py
new file mode 100644
index 0000000..4a6938c
--- /dev/null
+++ b/test/test_benchmark.py
@@ -0,0 +1,65 @@
+import os
+import subprocess
+import sys
+import unittest
+import tempfile
+
+from test.test_base import JSON
+
+
+class BenchmarkTests(unittest.TestCase):
+
+    def _do_test_benchmark(self, method="basic_parse", backend="python", multiple_values=True, extra_args=None):
+        # Use python backend to ensure multiple_values works
+        if multiple_values:
+            env = dict(os.environ)
+            env['IJSON_BACKEND'] = 'python'
+        # Ensure printing works on the subprocess in Windows
+        # by using utf-8 on its stdout
+        if sys.platform == 'win32':
+            env = dict(os.environ)
+            env['PYTHONIOENCODING'] = 'utf-8'
+        cmd = [sys.executable, '-m', 'ijson.benchmark', '-m', method, '-p', '', '-s', '1']
+        if multiple_values:
+            cmd.append('-M')
+        if extra_args:
+            cmd += extra_args
+        proc = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE, env=env)
+        out, err = proc.communicate()
+        status = proc.wait()
+        self.assertEqual(0, status, "out:\n%s\nerr:%s" % (out.decode('utf-8'), err.decode('utf-8')))
+
+    def _test_benchmark(self, method):
+        self._do_test_benchmark(method, "python")
+        self._do_test_benchmark(method, "python", extra_args=['-c'])
+        self._do_test_benchmark(method, "python", extra_args=['-a'])
+
+    def test_basic_parse(self):
+        self._test_benchmark('basic_parse')
+
+    def test_parse(self):
+        self._test_benchmark('parse')
+
+    def test_kvitems(self):
+        self._test_benchmark('kvitems')
+
+    def test_items(self):
+        self._test_benchmark('items')
+
+    def test_list(self):
+        self._do_test_benchmark(extra_args=['-l'])
+
+    def test_more_iterations(self):
+        self._do_test_benchmark(extra_args=['-I', '1'])
+        self._do_test_benchmark(extra_args=['-I', '2'])
+        self._do_test_benchmark(extra_args=['-I', '3'])
+
+    def test_input_files(self):
+        fd, fname = tempfile.mkstemp()
+        os.write(fd, JSON)
+        os.close(fd)
+        try:
+            self._do_test_benchmark(extra_args=[fname])
+            self._do_test_benchmark(extra_args=[fname, fname])
+        finally:
+            os.unlink(fname)