diff --git a/Include/internal/pycore_fileutils.h b/Include/internal/pycore_fileutils.h
index 13f86b01bbfe8f..74685a3e945459 100644
--- a/Include/internal/pycore_fileutils.h
+++ b/Include/internal/pycore_fileutils.h
@@ -70,8 +70,13 @@ extern PyObject* _Py_device_encoding(int);
 #   define _PY_WRITE_MAX INT_MAX
 #else
     /* write() should truncate the input to PY_SSIZE_T_MAX bytes,
-       but it's safer to do it ourself to have a portable behaviour */
-#   define _PY_READ_MAX  PY_SSIZE_T_MAX
+       but it's safer to do it ourself to have a portable behaviour
+
+       read() fills a PyBytes object, which has a capped size defined in
+       bytesobject.c. Prefer reading less data (meets the API spec) to causing
+       an overflow error.
+    */
+#   define _PY_READ_MAX  PY_SSIZE_T_MAX - 4096
 #   define _PY_WRITE_MAX PY_SSIZE_T_MAX
 #endif
 
diff --git a/Lib/test/test_os.py b/Lib/test/test_os.py
index f93937fb587386..6c0bb6a2cfc60c 100644
--- a/Lib/test/test_os.py
+++ b/Lib/test/test_os.py
@@ -234,10 +234,6 @@ def test_read(self):
             self.assertEqual(s, b"spam")
 
     @support.cpython_only
-    # Skip the test on 32-bit platforms: the number of bytes must fit in a
-    # Py_ssize_t type
-    @unittest.skipUnless(INT_MAX < PY_SSIZE_T_MAX,
-                         "needs INT_MAX < PY_SSIZE_T_MAX")
     @support.bigmemtest(size=INT_MAX + 10, memuse=1, dry_run=False)
     def test_large_read(self, size):
         self.addCleanup(os_helper.unlink, os_helper.TESTFN)
@@ -245,6 +241,9 @@ def test_large_read(self, size):
 
         # Issue #21932: Make sure that os.read() does not raise an
         # OverflowError for size larger than INT_MAX
+        #
+        # For 32 bit systems, it is expected the read doesn't read all the bytes
+        # but rather caps to a number it can reasonably read.
         with open(os_helper.TESTFN, "rb") as fp:
             data = os.read(fp.fileno(), size)
 
diff --git a/Misc/NEWS.d/next/Core and Builtins/2024-07-11-13-57-50.gh-issue-120754.C1HedA.rst b/Misc/NEWS.d/next/Core and Builtins/2024-07-11-13-57-50.gh-issue-120754.C1HedA.rst
new file mode 100644
index 00000000000000..47b50360225075
--- /dev/null
+++ b/Misc/NEWS.d/next/Core and Builtins/2024-07-11-13-57-50.gh-issue-120754.C1HedA.rst	
@@ -0,0 +1,3 @@
+``os.read()``  caps read size smaller than the max bytes object size to avoid
+getting an OverflowError that 'byte string is too large'. This makes it so the
+read call is attempted (although still may fail for other reasons).