diff --git a/litellm/llms/custom_httpx/http_handler.py b/litellm/llms/custom_httpx/http_handler.py
index 5ba96f87c53..4edaad38983 100644
--- a/litellm/llms/custom_httpx/http_handler.py
+++ b/litellm/llms/custom_httpx/http_handler.py
@@ -706,7 +706,7 @@ class _BrotliDecoder:
                 produced += len(piece)  # rebind-ok: see above
                 pending = b""  # rebind-ok: see above
                 yield piece
-                if self._inflate.can_accept_more_data():
+                if not piece or self._inflate.can_accept_more_data():
                     return
 
         return b"".join(drained())
diff --git a/tests/unit/llms/custom_httpx/test_http_handler.py b/tests/unit/llms/custom_httpx/test_http_handler.py
index d7c105bec74..1c1f6dc4c90 100644
--- a/tests/unit/llms/custom_httpx/test_http_handler.py
+++ b/tests/unit/llms/custom_httpx/test_http_handler.py
@@ -26,6 +26,7 @@ from litellm.llms.custom_httpx.http_handler import (
     HTTPHandler,
     HTTPResponseLimitError,
     MaskedHTTPStatusError,
+    _BrotliDecoder,
     _get_httpx_client,
     get_ssl_configuration,
 )
@@ -1934,6 +1935,27 @@ async def test_bounded_get_decodes_a_compressed_body_under_the_cap_and_rejects_a
         await handler.close()
 
 
+def test_brotli_decoder_stops_draining_when_process_returns_empty():
+    pieces: Final = iter((b"notes", b" more", b""))
+    calls: Final = []
+
+    class Inflate:
+        def process(self, data: bytes, output_buffer_limit: int) -> bytes:
+            calls.append(data)
+            if len(calls) > 8:
+                raise AssertionError("brotli drain loop did not stop after an empty process()")
+            return next(pieces, b"")
+
+        def can_accept_more_data(self) -> bool:
+            return False
+
+        def is_finished(self) -> bool:
+            return True
+
+    assert _BrotliDecoder(Inflate()).decode(b"wire", max_output=1024) == b"notes more"
+    assert calls == [b"wire", b"", b""]
+
+
 @pytest.mark.asyncio
 async def test_bounded_get_refuses_an_encoding_it_cannot_decode_under_the_cap(respx_mock, monkeypatch):
     monkeypatch.setenv("DISABLE_AIOHTTP_TRANSPORT", "True")