diff --git a/python/sglang/srt/mem_cache/cp_shared_l2_pool.py b/python/sglang/srt/mem_cache/cp_shared_l2_pool.py
index 058a6e699..592907990 100644
--- a/python/sglang/srt/mem_cache/cp_shared_l2_pool.py
+++ b/python/sglang/srt/mem_cache/cp_shared_l2_pool.py
@@ -804,6 +804,30 @@ class CpSharedL2PageAllocator:
             self._stats["cp_shared_l2_objects_evicted"] += 1
         return dropped
 
+    def mark_object_committed(self, object_key: str) -> bool:
+        """B1 commit: mark an object committed directly, bypassing the per-(rank,
+        layer, payload) commit_layer quorum.
+
+        Under the B1 collective-free allocator the all-ranks-done consensus is the
+        writing_check ReduceOp.MIN frontier (the same barrier that releases the
+        node's write lock), NOT a per-layer all-gather. When that frontier commits a
+        node, every CP rank calls this with the same object_key, so the committed set
+        transitions rank-uniformly (and feeds placement_digest). Does not populate
+        _commits_by_object -- the per-layer quorum (_has_full_commit) is unused under
+        B1; is_committed / release / split_committed_object read _committed_objects,
+        which this maintains. Returns True on the first commit, False if already
+        committed. Raises on an unknown object (fail-loud: the node's reservation
+        must outlive its write lock).
+        """
+        self._validate_object_key(object_key)
+        if object_key not in self._ranges_by_object:
+            raise ValueError(f"cannot commit unknown object {object_key!r}")
+        if object_key in self._committed_objects:
+            return False
+        self._committed_objects.add(object_key)
+        self._stats["cp_shared_l2_objects_committed"] += 1
+        return True
+
     def split_committed_object(
         self,
         object_key: str,
diff --git a/test/registered/unit/mem_cache/test_cp_shared_l2_pool.py b/test/registered/unit/mem_cache/test_cp_shared_l2_pool.py
index 1a26c5153..e0abce0a1 100644
--- a/test/registered/unit/mem_cache/test_cp_shared_l2_pool.py
+++ b/test/registered/unit/mem_cache/test_cp_shared_l2_pool.py
@@ -2784,5 +2784,83 @@ class TestCpSharedL2PlacementDigest(unittest.TestCase):
         self.assertNotEqual(a.placement_digest(), b.placement_digest())
 
 
+class TestCpSharedL2MarkObjectCommitted(unittest.TestCase):
+    """B1 commit path (2b.1b): mark_object_committed replaces the per-(rank,layer,
+    payload) quorum -- the writing_check MIN frontier is the consensus. Locks in
+    MF4 (split stays coherent for a B1-committed object with no per-layer commits)."""
+
+    def make_allocator(self, pages=16):
+        return _cp_shared_l2_pool.CpSharedL2PageAllocator(
+            pages_per_payload={PAYLOAD_TARGET_KV: pages},
+            slab_ids_by_payload={PAYLOAD_TARGET_KV: 10},
+            expected_ranks=(0, 1, 2, 3),
+            expected_layers=(0, 1),
+            required_payloads=(PAYLOAD_TARGET_KV,),
+        )
+
+    def test_mark_commits_without_per_layer_quorum(self):
+        a = self.make_allocator()
+        a.reserve("n", PAYLOAD_TARGET_KV, 3)
+        self.assertFalse(a.is_committed("n"))
+        self.assertTrue(a.mark_object_committed("n"))
+        self.assertTrue(a.is_committed("n"))
+
+    def test_mark_is_idempotent(self):
+        a = self.make_allocator()
+        a.reserve("n", PAYLOAD_TARGET_KV, 2)
+        self.assertTrue(a.mark_object_committed("n"))
+        self.assertFalse(a.mark_object_committed("n"))
+
+    def test_mark_unknown_object_raises(self):
+        a = self.make_allocator()
+        with self.assertRaises(ValueError):
+            a.mark_object_committed("never-reserved")
+
+    def test_mark_bumps_committed_stat(self):
+        a = self.make_allocator()
+        a.reserve("n", PAYLOAD_TARGET_KV, 2)
+        before = a.stats()["cp_shared_l2_objects_committed"]
+        a.mark_object_committed("n")
+        self.assertEqual(a.stats()["cp_shared_l2_objects_committed"], before + 1)
+
+    def test_mark_changes_placement_digest(self):
+        # committed set is part of the digest, so commit transitions are visible to
+        # the cross-rank assert; and two ranks committing in lockstep stay equal.
+        a, b = self.make_allocator(), self.make_allocator()
+        a.reserve("n", PAYLOAD_TARGET_KV, 3)
+        b.reserve("n", PAYLOAD_TARGET_KV, 3)
+        self.assertEqual(a.placement_digest(), b.placement_digest())
+        a.mark_object_committed("n")
+        self.assertNotEqual(a.placement_digest(), b.placement_digest())
+        b.mark_object_committed("n")
+        self.assertEqual(a.placement_digest(), b.placement_digest())
+
+    def test_split_of_marked_committed_object_is_coherent(self):
+        # MF4: split a B1-committed object (no per-layer commit facts) -> children
+        # are committed, pages preserved, no error from the empty _commits_by_object.
+        a = self.make_allocator()
+        a.reserve("child", PAYLOAD_TARGET_KV, 6)
+        a.mark_object_committed("child")
+        parent_ranges, child_ranges = a.split_committed_object(
+            "child",
+            split_pages_by_payload={PAYLOAD_TARGET_KV: 2},
+            parent_object_key="parent",
+            child_object_key="child",
+        )
+        self.assertTrue(a.is_committed("parent"))
+        self.assertTrue(a.is_committed("child"))
+        self.assertEqual(parent_ranges[PAYLOAD_TARGET_KV].num_pages, 2)
+        self.assertEqual(child_ranges[PAYLOAD_TARGET_KV].num_pages, 4)
+
+    def test_release_after_mark_frees_pages(self):
+        a = self.make_allocator(pages=8)
+        free0 = a.free_pages(PAYLOAD_TARGET_KV)
+        a.reserve("n", PAYLOAD_TARGET_KV, 4)
+        a.mark_object_committed("n")
+        self.assertTrue(a.release("n"))
+        self.assertEqual(a.free_pages(PAYLOAD_TARGET_KV), free0)
+        self.assertFalse(a.is_committed("n"))
+
+
 if __name__ == "__main__":
     unittest.main()