diff --git a/python/sglang/srt/mem_cache/pure_swa_radix_cache.py b/python/sglang/srt/mem_cache/pure_swa_radix_cache.py
index 6dbed5b86..0636d0c13 100644
--- a/python/sglang/srt/mem_cache/pure_swa_radix_cache.py
+++ b/python/sglang/srt/mem_cache/pure_swa_radix_cache.py
@@ -129,8 +129,8 @@ class PureSWARadixCache(RadixCache):
                 self.token_to_kv_pool_allocator.free(
                     kv_indices[old_prefix_len:free_end]
                 )
-            alive_start = max(swa_evicted_seqlen, old_prefix_len)
-            if swa_evicted_seqlen > 0 and alive_start < keys_len:
+            alive_start = max(swa_evicted_seqlen, free_end)
+            if alive_start < keys_len:
                 self.token_to_kv_pool_allocator.free(kv_indices[alive_start:keys_len])
 
         self.token_to_kv_pool_allocator.free(kv_indices[keys_len:])
diff --git a/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py b/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py
new file mode 100644
index 000000000..b9f108dd1
--- /dev/null
+++ b/test/registered/unit/mem_cache/test_pure_swa_radix_cache.py
@@ -0,0 +1,57 @@
+"""Unit tests for all-SWA RadixCache release semantics."""
+
+import unittest
+from types import SimpleNamespace
+
+import torch
+
+from sglang.srt.mem_cache.pure_swa_radix_cache import PureSWARadixCache
+from sglang.test.ci.ci_register import register_cpu_ci
+from sglang.test.test_utils import CustomTestCase
+
+register_cpu_ci(est_time=10, suite="base-a-test-cpu")
+
+
+class _FakeAllocator:
+    def __init__(self):
+        self.freed = []
+
+    def free(self, indices):
+        if indices.numel() == 0:
+            return
+        self.freed.append(indices.detach().cpu().tolist())
+
+
+class TestPureSWARadixCache(CustomTestCase):
+    def test_no_insert_frees_window_after_evict_floor_before_swa_eviction(self):
+        allocator = _FakeAllocator()
+        cache = PureSWARadixCache.__new__(PureSWARadixCache)
+        cache.disable_finished_insert = False
+        cache.disable = False
+        cache.is_eagle = False
+        cache.page_size = 1
+        cache.req_to_token_pool = SimpleNamespace(
+            req_to_token=torch.arange(8, dtype=torch.int64).unsqueeze(0)
+        )
+        cache.token_to_kv_pool_allocator = allocator
+        req = SimpleNamespace(
+            origin_input_ids=list(range(8)),
+            output_ids=[],
+            extra_key=None,
+            cache_salt=None,
+            last_node=None,
+            kv=SimpleNamespace(
+                req_pool_idx=0,
+                cache_protected_len=0,
+                swa_evict_floor=4,
+                swa_evicted_seqlen=0,
+            ),
+        )
+
+        cache.cache_finished_req(req, is_insert=False, kv_len_to_handle=8)
+
+        self.assertEqual(allocator.freed, [[0, 1, 2, 3], [4, 5, 6, 7]])
+
+
+if __name__ == "__main__":
+    unittest.main()