diff --git a/packages/http/httpx/kiota_http/middleware/retry_handler.py b/packages/http/httpx/kiota_http/middleware/retry_handler.py
index fa105fc..2c06ee2 100644
--- a/packages/http/httpx/kiota_http/middleware/retry_handler.py
+++ b/packages/http/httpx/kiota_http/middleware/retry_handler.py
@@ -95,6 +95,8 @@ async def send(self, request: httpx.Request, transport: httpx.AsyncBaseTransport
             # and status code
             should_retry = self.should_retry(request, current_options, response)
             if all([should_retry, retry_valid, delay < RetryHandlerOption.MAX_DELAY]):
+                # release the discarded response's connection before retrying
+                await response.aclose()
                 await asyncio.sleep(delay)
                 # increment the count for retries
                 retry_count += 1
diff --git a/packages/http/httpx/tests/middleware_tests/test_retry_handler.py b/packages/http/httpx/tests/middleware_tests/test_retry_handler.py
index 9f2edda..e45a12b 100644
--- a/packages/http/httpx/tests/middleware_tests/test_retry_handler.py
+++ b/packages/http/httpx/tests/middleware_tests/test_retry_handler.py
@@ -298,3 +298,32 @@ def request_handler(request: httpx.Request):
     assert resp.status_code == GATEWAY_TIMEOUT
     assert 'request_2' in resp.request.headers
     assert resp.request.headers[RETRY_ATTEMPT] == '2'
+
+
+@pytest.mark.asyncio
+async def test_retried_responses_are_closed():
+    """Discarded retry responses are closed so their connection is released."""
+    responses = []
+
+    async def body():
+        yield b'{}'
+
+    statuses = [TOO_MANY_REQUESTS, TOO_MANY_REQUESTS, 200]
+
+    def request_handler(request: httpx.Request):
+        status = statuses.pop(0)
+        headers = {RETRY_AFTER: '0'} if status == TOO_MANY_REQUESTS else {}
+        response = httpx.Response(status, headers=headers, content=body())
+        responses.append(response)
+        return response
+
+    handler = RetryHandler(RetryHandlerOption(delay=0, max_retries=2))
+    request = httpx.Request('GET', BASE_URL)
+    mock_transport = httpx.MockTransport(request_handler)
+    resp = await handler.send(request, mock_transport)
+
+    assert resp.status_code == 200
+    # the two retried responses are closed, releasing their connections
+    assert responses[0].is_closed and responses[1].is_closed
+    # the returned response is left open for the caller to read
+    assert not responses[2].is_closed