AstrBotDevs
diff --git a/astrbot/core/provider/sources/openai_source.py b/astrbot/core/provider/sources/openai_source.py
@@ -377,6 +377,9 @@ def __init__(self, provider_config, provider_settings) -> None:
                 default_headers=self.custom_headers,
                 base_url=provider_config.get("api_base", ""),
                 timeout=self.timeout,
+                # Retry is handled by retry_provider_request(); disable the
+                # SDK built-in retry to avoid stacking request attempts.
+                max_retries=0,
                 http_client=self._create_http_client(provider_config),
             )
         else:
@@ -386,6 +389,9 @@ def __init__(self, provider_config, provider_settings) -> None:
                 base_url=provider_config.get("api_base", None),
                 default_headers=self.custom_headers,
                 timeout=self.timeout,
+                # Retry is handled by retry_provider_request(); disable the
+                # SDK built-in retry to avoid stacking request attempts.
+                max_retries=0,
                 http_client=self._create_http_client(provider_config),
             )
 

diff --git a/tests/test_openai_source.py b/tests/test_openai_source.py
@@ -5,6 +5,7 @@
 
 import httpx
 import pytest
+from openai import AsyncAzureOpenAI, AsyncOpenAI
 from openai.types.chat.chat_completion import ChatCompletion
 from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
 from PIL import Image as PILImage
@@ -60,6 +61,54 @@ def _make_groq_provider(overrides: dict | None = None) -> ProviderGroq:
     )
 
 
+@pytest.mark.parametrize(
+    ("overrides", "expected_client"),
+    [
+        ({}, AsyncOpenAI),
+        ({"api_version": "2024-02-01"}, AsyncAzureOpenAI),
+    ],
+)
+@pytest.mark.asyncio
+async def test_provider_client_disables_sdk_builtin_retries(overrides, expected_client):
+    provider = _make_provider(overrides)
+    try:
+        assert isinstance(provider.client, expected_client)
+        assert provider.client.max_retries == 0
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_query_attempts_exactly_request_max_retries_times(monkeypatch):
+    monkeypatch.setattr(request_retry, "REQUEST_RETRY_WAIT_MIN_S", 0)
+    monkeypatch.setattr(request_retry, "REQUEST_RETRY_WAIT_MAX_S", 0)
+
+    provider = _make_provider()
+    try:
+        calls = 0
+
+        async def failing_create(**kwargs):
+            nonlocal calls
+            calls += 1
+            raise httpx.ConnectError("temporary connection failure")
+
+        monkeypatch.setattr(provider.client.chat.completions, "create", failing_create)
+
+        with pytest.raises(httpx.ConnectError):
+            await provider._query(
+                payloads={
+                    "model": "gpt-4o-mini",
+                    "messages": [{"role": "user", "content": "hi"}],
+                },
+                tools=None,
+                request_max_retries=2,
+            )
+
+        assert calls == 2
+    finally:
+        await provider.terminate()
+
+
 def test_create_http_client_uses_openai_httpx_module(monkeypatch):
     captured: dict[str, object] = {}
     fake_httpx_module = object()