diff --git a/pyproject.toml b/pyproject.toml
index 1fad8a6..edbe330 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -97,7 +97,6 @@ markers = [
 # Warnings should only be emitted when being specifically tested
 filterwarnings = [
     "error",
-    "ignore:.*the async API is not yet stable:FutureWarning"
 ]
 # Capture log info from network client libraries
 log_format = "%(asctime)s %(levelname)s %(message)s"
diff --git a/src/lmstudio/_ws_impl.py b/src/lmstudio/_ws_impl.py
index 80dba6f..2565f29 100644
--- a/src/lmstudio/_ws_impl.py
+++ b/src/lmstudio/_ws_impl.py
@@ -11,7 +11,10 @@
 
 import asyncio
 
-from concurrent.futures import Future as SyncFuture
+
+# Python 3.10 compatibility: use concurrent.futures.TimeoutError instead of the builtin
+# In 3.11+, these are the same type, in 3.10 futures have their own timeout exception
+from concurrent.futures import Future as SyncFuture, TimeoutError as SyncFutureTimeout
 from contextlib import AsyncExitStack, contextmanager
 from functools import partial
 from typing import (
@@ -47,6 +50,12 @@
 # and omits the generalised features that the SDK doesn't need)
 T = TypeVar("T")
 
+__all__ = [
+    "SyncFutureTimeout",
+    "AsyncTaskManager",
+    "AsyncWebsocketHandler",
+]
+
 
 class AsyncTaskManager:
     def __init__(self, *, on_activation: Callable[[], Any] | None = None) -> None:
@@ -429,7 +438,7 @@ def _rx_queue_get_threadsafe(self, rx_queue: RxQueue, timeout: float | None) ->
         future = self._task_manager.run_coroutine_threadsafe(rx_queue.get())
         try:
             return future.result(timeout)
-        except TimeoutError:
+        except SyncFutureTimeout:
             future.cancel()
             raise
 
diff --git a/src/lmstudio/async_api.py b/src/lmstudio/async_api.py
index 25605a4..ffaedcd 100644
--- a/src/lmstudio/async_api.py
+++ b/src/lmstudio/async_api.py
@@ -1,7 +1,6 @@
 """Async I/O protocol implementation for the LM Studio remote access API."""
 
 import asyncio
-import warnings
 
 from abc import abstractmethod
 from contextlib import AsyncExitStack, asynccontextmanager
@@ -109,8 +108,8 @@
 # and similar tasks is published from `json_api`.
 # Bypassing the high level API, and working more
 # directly with the underlying websocket(s) is
-# supported (hence the public names), but they're
-# not exported via the top-level `lmstudio` API.
+# not supported due to the complexity of the task
+# management details (hence the private names).
 __all__ = [
     "AnyAsyncDownloadedModel",
     "AsyncClient",
@@ -215,7 +214,7 @@ async def receive_result(self) -> Any:
         return self._rpc.handle_rx_message(message)
 
 
-class AsyncLMStudioWebsocket(LMStudioWebsocket[AsyncWebSocketSession]):
+class _AsyncLMStudioWebsocket(LMStudioWebsocket[AsyncWebSocketSession]):
     """Asynchronous websocket client that handles demultiplexing of reply messages."""
 
     def __init__(
@@ -331,7 +330,7 @@ async def remote_call(
             return await rpc.receive_result()
 
 
-class AsyncSession(ClientSession["AsyncClient", AsyncLMStudioWebsocket]):
+class _AsyncSession(ClientSession["AsyncClient", _AsyncLMStudioWebsocket]):
     """Async client session interfaces applicable to all API namespaces."""
 
     def __init__(self, client: "AsyncClient") -> None:
@@ -354,7 +353,7 @@ async def __aexit__(self, *args: Any) -> None:
         await self.disconnect()
 
     @sdk_public_api_async()
-    async def connect(self) -> AsyncLMStudioWebsocket:
+    async def connect(self) -> _AsyncLMStudioWebsocket:
         """Connect the client session."""
         self._fail_if_connected("Attempted to connect already connected session")
         api_host = self._client.api_host
@@ -367,7 +366,7 @@ async def connect(self) -> AsyncLMStudioWebsocket:
         resources = self._resource_manager
         client = self._client
         self._lmsws = lmsws = await resources.enter_async_context(
-            AsyncLMStudioWebsocket(
+            _AsyncLMStudioWebsocket(
                 client._task_manager, session_url, client._auth_details
             )
         )
@@ -411,7 +410,7 @@ async def remote_call(
 
 
 TAsyncSessionModel = TypeVar(
-    "TAsyncSessionModel", bound="AsyncSessionModel[Any, Any, Any, Any]"
+    "TAsyncSessionModel", bound="_AsyncSessionModel[Any, Any, Any, Any]"
 )
 TAsyncModelHandle = TypeVar("TAsyncModelHandle", bound="AsyncModelHandle[Any]")
 
@@ -467,7 +466,7 @@ async def model(
 class AsyncDownloadedEmbeddingModel(
     AsyncDownloadedModel[
         EmbeddingModelInfo,
-        "AsyncSessionEmbedding",
+        "_AsyncSessionEmbedding",
         EmbeddingLoadModelConfig,
         EmbeddingLoadModelConfigDict,
         "AsyncEmbeddingModel",
@@ -476,7 +475,7 @@ class AsyncDownloadedEmbeddingModel(
     """Asynchronous download listing for an embedding model."""
 
     def __init__(
-        self, model_info: DictObject, session: "AsyncSessionEmbedding"
+        self, model_info: DictObject, session: "_AsyncSessionEmbedding"
     ) -> None:
         """Initialize downloaded embedding model details."""
         super().__init__(EmbeddingModelInfo, model_info, session)
@@ -485,7 +484,7 @@ def __init__(
 class AsyncDownloadedLlm(
     AsyncDownloadedModel[
         LlmInfo,
-        "AsyncSessionLlm",
+        "_AsyncSessionLlm",
         LlmLoadModelConfig,
         LlmLoadModelConfigDict,
         "AsyncLLM",
@@ -493,7 +492,7 @@ class AsyncDownloadedLlm(
 ):
     """Asynchronous ownload listing for an LLM."""
 
-    def __init__(self, model_info: DictObject, session: "AsyncSessionLlm") -> None:
+    def __init__(self, model_info: DictObject, session: "_AsyncSessionLlm") -> None:
         """Initialize downloaded embedding model details."""
         super().__init__(LlmInfo, model_info, session)
 
@@ -501,7 +500,7 @@ def __init__(self, model_info: DictObject, session: "AsyncSessionLlm") -> None:
 AnyAsyncDownloadedModel: TypeAlias = AsyncDownloadedModel[Any, Any, Any, Any, Any]
 
 
-class AsyncSessionSystem(AsyncSession):
+class _AsyncSessionSystem(_AsyncSession):
     """Async client session for the system namespace."""
 
     API_NAMESPACE = "system"
@@ -531,7 +530,7 @@ def _process_download_listing(
         )
 
 
-class _AsyncSessionFiles(AsyncSession):
+class _AsyncSessionFiles(_AsyncSession):
     """Async client session for the files namespace."""
 
     API_NAMESPACE = "files"
@@ -562,7 +561,7 @@ async def prepare_image(
         return await self._fetch_file_handle(file_data)
 
 
-class AsyncModelDownloadOption(ModelDownloadOptionBase[AsyncSession]):
+class AsyncModelDownloadOption(ModelDownloadOptionBase[_AsyncSession]):
     """A single download option for a model search result."""
 
     @sdk_public_api_async()
@@ -577,10 +576,10 @@ async def download(
             return await channel.wait_for_result()
 
 
-class AsyncAvailableModel(AvailableModelBase[AsyncSession]):
+class AsyncAvailableModel(AvailableModelBase[_AsyncSession]):
     """A model available for download from the model repository."""
 
-    _session: AsyncSession
+    _session: _AsyncSession
 
     @sdk_public_api_async()
     async def get_download_options(
@@ -595,7 +594,7 @@ async def get_download_options(
         return final
 
 
-class AsyncSessionRepository(AsyncSession):
+class _AsyncSessionRepository(_AsyncSession):
     """Async client session for the repository namespace."""
 
     API_NAMESPACE = "repository"
@@ -616,8 +615,8 @@ async def search_models(
 TAsyncDownloadedModel = TypeVar("TAsyncDownloadedModel", bound=AnyAsyncDownloadedModel)
 
 
-class AsyncSessionModel(
-    AsyncSession,
+class _AsyncSessionModel(
+    _AsyncSession,
     Generic[
         TAsyncModelHandle,
         TLoadConfig,
@@ -630,7 +629,7 @@ class AsyncSessionModel(
     _API_TYPES: Type[ModelSessionTypes[TLoadConfig]]
 
     @property
-    def _system_session(self) -> AsyncSessionSystem:
+    def _system_session(self) -> _AsyncSessionSystem:
         return self._client.system
 
     @property
@@ -922,8 +921,8 @@ async def cancel(self) -> None:
             await self._channel.cancel()
 
 
-class AsyncSessionLlm(
-    AsyncSessionModel[
+class _AsyncSessionLlm(
+    _AsyncSessionModel[
         "AsyncLLM",
         LlmLoadModelConfig,
         LlmLoadModelConfigDict,
@@ -1028,8 +1027,8 @@ async def _apply_prompt_template(
         return response.get("formatted", "") if response else ""
 
 
-class AsyncSessionEmbedding(
-    AsyncSessionModel[
+class _AsyncSessionEmbedding(
+    _AsyncSessionModel[
         "AsyncEmbeddingModel",
         EmbeddingLoadModelConfig,
         EmbeddingLoadModelConfigDict,
@@ -1115,7 +1114,7 @@ async def get_context_length(self) -> int:
 AnyAsyncModel: TypeAlias = AsyncModelHandle[Any]
 
 
-class AsyncLLM(AsyncModelHandle[AsyncSessionLlm]):
+class AsyncLLM(AsyncModelHandle[_AsyncSessionLlm]):
     """Reference to a loaded LLM model."""
 
     @sdk_public_api_async()
@@ -1258,7 +1257,7 @@ async def apply_prompt_template(
         )
 
 
-class AsyncEmbeddingModel(AsyncModelHandle[AsyncSessionEmbedding]):
+class AsyncEmbeddingModel(AsyncModelHandle[_AsyncSessionEmbedding]):
     """Reference to a loaded embedding model."""
 
     # Alas, type hints don't properly support distinguishing str vs Iterable[str]:
@@ -1271,11 +1270,7 @@ async def embed(
         return await self._session._embed(self.identifier, input)
 
 
-TAsyncSession = TypeVar("TAsyncSession", bound=AsyncSession)
-
-_ASYNC_API_STABILITY_WARNING = """\
-Note the async API is not yet stable and is expected to change in future releases
-"""
+TAsyncSession = TypeVar("TAsyncSession", bound=_AsyncSession)
 
 
 class AsyncClient(ClientBase):
@@ -1283,12 +1278,9 @@ class AsyncClient(ClientBase):
 
     def __init__(self, api_host: str | None = None) -> None:
         """Initialize API client."""
-        # Warn about the async API stability, since we expect it to change
-        # (in particular, accepting coroutine functions as callbacks)
-        warnings.warn(_ASYNC_API_STABILITY_WARNING, FutureWarning)
         super().__init__(api_host)
         self._resources = AsyncExitStack()
-        self._sessions: dict[str, AsyncSession] = {}
+        self._sessions: dict[str, _AsyncSession] = {}
         self._task_manager = AsyncTaskManager()
         # Unlike the sync API, we don't support GC-based resource
         # management in the async API. Structured concurrency
@@ -1301,12 +1293,12 @@ def __init__(self, api_host: str | None = None) -> None:
     # TODO: revisit lazy connections given the task manager implementation
     #       (for example, eagerly start tasks for all sessions, and lazily
     #       trigger events that allow them to initiate their connection)
-    _ALL_SESSIONS: tuple[Type[AsyncSession], ...] = (
-        AsyncSessionEmbedding,
+    _ALL_SESSIONS: tuple[Type[_AsyncSession], ...] = (
+        _AsyncSessionEmbedding,
         _AsyncSessionFiles,
-        AsyncSessionLlm,
-        AsyncSessionRepository,
-        AsyncSessionSystem,
+        _AsyncSessionLlm,
+        _AsyncSessionRepository,
+        _AsyncSessionSystem,
     )
 
     async def __aenter__(self) -> Self:
@@ -1342,20 +1334,20 @@ def _get_session(self, cls: Type[TAsyncSession]) -> TAsyncSession:
 
     @property
     @sdk_public_api()
-    def llm(self) -> AsyncSessionLlm:
+    def llm(self) -> _AsyncSessionLlm:
         """Return the LLM API client session."""
-        return self._get_session(AsyncSessionLlm)
+        return self._get_session(_AsyncSessionLlm)
 
     @property
     @sdk_public_api()
-    def embedding(self) -> AsyncSessionEmbedding:
+    def embedding(self) -> _AsyncSessionEmbedding:
         """Return the embedding model API client session."""
-        return self._get_session(AsyncSessionEmbedding)
+        return self._get_session(_AsyncSessionEmbedding)
 
     @property
-    def system(self) -> AsyncSessionSystem:
+    def system(self) -> _AsyncSessionSystem:
         """Return the system API client session."""
-        return self._get_session(AsyncSessionSystem)
+        return self._get_session(_AsyncSessionSystem)
 
     @property
     def files(self) -> _AsyncSessionFiles:
@@ -1363,9 +1355,9 @@ def files(self) -> _AsyncSessionFiles:
         return self._get_session(_AsyncSessionFiles)
 
     @property
-    def repository(self) -> AsyncSessionRepository:
+    def repository(self) -> _AsyncSessionRepository:
         """Return the repository API client session."""
-        return self._get_session(AsyncSessionRepository)
+        return self._get_session(_AsyncSessionRepository)
 
     # Convenience methods
     # Not yet implemented (server API only supports the same file types as prepare_image)
diff --git a/src/lmstudio/plugin/cli.py b/src/lmstudio/plugin/cli.py
index a528528..5a7a9fe 100644
--- a/src/lmstudio/plugin/cli.py
+++ b/src/lmstudio/plugin/cli.py
@@ -44,9 +44,6 @@ def main(argv: Sequence[str] | None = None) -> int:
     warnings.filterwarnings(
         "ignore", ".*the plugin API is not yet stable", FutureWarning
     )
-    warnings.filterwarnings(
-        "ignore", ".*the async API is not yet stable", FutureWarning
-    )
     log_level = logging.DEBUG if args.debug else logging.INFO
     logging.basicConfig(level=log_level)
     if not args.dev:
diff --git a/src/lmstudio/plugin/hooks/common.py b/src/lmstudio/plugin/hooks/common.py
index 8ed4d35..80289d0 100644
--- a/src/lmstudio/plugin/hooks/common.py
+++ b/src/lmstudio/plugin/hooks/common.py
@@ -18,7 +18,7 @@
 
 from anyio import move_on_after
 
-from ...async_api import AsyncSession
+from ...async_api import _AsyncSession
 from ...schemas import DictObject
 from ..._sdk_models import (
     # TODO: Define aliases at schema generation time
@@ -32,13 +32,13 @@
 
 # Available as lmstudio.plugin.hooks.*
 __all__ = [
-    "AsyncSessionPlugins",
+    "_AsyncSessionPlugins",
     "TPluginConfigSchema",
     "TGlobalConfigSchema",
 ]
 
 
-class AsyncSessionPlugins(AsyncSession):
+class _AsyncSessionPlugins(_AsyncSession):
     """Async client session for the plugins namespace."""
 
     API_NAMESPACE = "plugins"
@@ -63,7 +63,7 @@ class HookController(Generic[TPluginRequest, TPluginConfigSchema, TGlobalConfigS
 
     def __init__(
         self,
-        session: AsyncSessionPlugins,
+        session: _AsyncSessionPlugins,
         request: TPluginRequest,
         plugin_config_schema: type[TPluginConfigSchema],
         global_config_schema: type[TGlobalConfigSchema],
diff --git a/src/lmstudio/plugin/hooks/prompt_preprocessor.py b/src/lmstudio/plugin/hooks/prompt_preprocessor.py
index 73a4910..a1a3175 100644
--- a/src/lmstudio/plugin/hooks/prompt_preprocessor.py
+++ b/src/lmstudio/plugin/hooks/prompt_preprocessor.py
@@ -47,7 +47,7 @@
 )
 from ..config_schemas import BaseConfigSchema
 from .common import (
-    AsyncSessionPlugins,
+    _AsyncSessionPlugins,
     HookController,
     SendMessageCallback,
     ServerRequestError,
@@ -129,7 +129,7 @@ class PromptPreprocessorController(
 
     def __init__(
         self,
-        session: AsyncSessionPlugins,
+        session: _AsyncSessionPlugins,
         request: PromptPreprocessingRequest,
         plugin_config_schema: type[TPluginConfigSchema],
         global_config_schema: type[TGlobalConfigSchema],
@@ -210,7 +210,7 @@ def __post_init__(self) -> None:
         self._abort_events: dict[str, asyncio.Event] = {}
 
     async def process_requests(
-        self, session: AsyncSessionPlugins, notify_ready: Callable[[], Any]
+        self, session: _AsyncSessionPlugins, notify_ready: Callable[[], Any]
     ) -> None:
         """Create plugin channel and wait for server requests."""
         logger = self._logger
@@ -376,7 +376,7 @@ async def run_prompt_preprocessor(
     hook_impl: PromptPreprocessorHook,
     plugin_config_schema: type[BaseConfigSchema],
     global_config_schema: type[BaseConfigSchema],
-    session: AsyncSessionPlugins,
+    session: _AsyncSessionPlugins,
     notify_ready: Callable[[], Any],
 ) -> None:
     """Accept prompt preprocessing requests."""
diff --git a/src/lmstudio/plugin/hooks/token_generator.py b/src/lmstudio/plugin/hooks/token_generator.py
index 51bb5ce..76d53b8 100644
--- a/src/lmstudio/plugin/hooks/token_generator.py
+++ b/src/lmstudio/plugin/hooks/token_generator.py
@@ -8,7 +8,7 @@
 
 from ..config_schemas import BaseConfigSchema
 from .common import (
-    AsyncSessionPlugins,
+    _AsyncSessionPlugins,
     HookController,
     TPluginConfigSchema,
     TGlobalConfigSchema,
@@ -36,7 +36,7 @@ async def run_token_generator(
     hook_impl: TokenGeneratorHook,
     plugin_config_schema: type[BaseConfigSchema],
     global_config_schema: type[BaseConfigSchema],
-    session: AsyncSessionPlugins,
+    session: _AsyncSessionPlugins,
     notify_ready: Callable[[], Any],
 ) -> None:
     """Accept token generation requests."""
diff --git a/src/lmstudio/plugin/hooks/tools_provider.py b/src/lmstudio/plugin/hooks/tools_provider.py
index e43b096..67d2e3b 100644
--- a/src/lmstudio/plugin/hooks/tools_provider.py
+++ b/src/lmstudio/plugin/hooks/tools_provider.py
@@ -13,7 +13,7 @@
 
 from ..config_schemas import BaseConfigSchema
 from .common import (
-    AsyncSessionPlugins,
+    _AsyncSessionPlugins,
     HookController,
     TPluginConfigSchema,
     TGlobalConfigSchema,
@@ -43,7 +43,7 @@ async def run_tools_provider(
     hook_impl: ToolsProviderHook,
     plugin_config_schema: type[BaseConfigSchema],
     global_config_schema: type[BaseConfigSchema],
-    session: AsyncSessionPlugins,
+    session: _AsyncSessionPlugins,
     notify_ready: Callable[[], Any],
 ) -> None:
     """Accept tools provider session requests."""
diff --git a/src/lmstudio/plugin/runner.py b/src/lmstudio/plugin/runner.py
index fa81ba7..8511a65 100644
--- a/src/lmstudio/plugin/runner.py
+++ b/src/lmstudio/plugin/runner.py
@@ -27,7 +27,7 @@
 from .sdk_api import LMStudioPluginInitError, LMStudioPluginRuntimeError
 from .config_schemas import BaseConfigSchema
 from .hooks import (
-    AsyncSessionPlugins,
+    _AsyncSessionPlugins,
     TPluginConfigSchema,
     TGlobalConfigSchema,
     run_prompt_preprocessor,
@@ -55,7 +55,7 @@
         THookImpl,
         type[TPluginConfigSchema],
         type[TGlobalConfigSchema],
-        AsyncSessionPlugins,
+        _AsyncSessionPlugins,
         ReadyCallback,
     ],
     Awaitable[Any],
@@ -100,7 +100,7 @@ def __init__(
 
     _ALL_SESSIONS = (
         # Plugin controller access all runs through a dedicated endpoint
-        AsyncSessionPlugins,
+        _AsyncSessionPlugins,
     )
 
     def _create_auth_message(self) -> DictObject:
@@ -111,9 +111,9 @@ def _create_auth_message(self) -> DictObject:
         return self._format_auth_message(self._client_id, self._client_key)
 
     @property
-    def plugins(self) -> AsyncSessionPlugins:
+    def plugins(self) -> _AsyncSessionPlugins:
         """Return the plugins API client session."""
-        return self._get_session(AsyncSessionPlugins)
+        return self._get_session(_AsyncSessionPlugins)
 
     async def _run_hook_impl(
         self,
diff --git a/src/lmstudio/sync_api.py b/src/lmstudio/sync_api.py
index b168eb8..c4c9ae6 100644
--- a/src/lmstudio/sync_api.py
+++ b/src/lmstudio/sync_api.py
@@ -110,6 +110,7 @@
     _model_spec_to_api_dict,
     _redact_json,
 )
+from ._ws_impl import SyncFutureTimeout
 from ._ws_thread import AsyncWebsocketThread, SyncToAsyncWebsocketBridge
 from ._kv_config import TLoadConfig, TLoadConfigDict, parse_server_config
 from ._sdk_models import (
@@ -228,7 +229,7 @@ def rx_stream(
                 # (we can't easily suppress the SDK's own frames for iterators)
                 try:
                     message = self._get_message(self.timeout)
-                except TimeoutError:
+                except SyncFutureTimeout:
                     raise LMStudioTimeoutError from None
                 if message is None:
                     raise LMStudioWebsocketError("Client unexpectedly disconnected.")
@@ -284,7 +285,7 @@ def receive_result(self) -> Any:
         """Receive call response on the receive queue."""
         try:
             message = self._get_message(self.timeout)
-        except TimeoutError:
+        except SyncFutureTimeout:
             raise LMStudioTimeoutError from None
         if message is None:
             raise LMStudioWebsocketError("Client unexpectedly disconnected.")
diff --git a/tests/test_session_errors.py b/tests/test_session_errors.py
index ec15b3e..f4e3e2a 100644
--- a/tests/test_session_errors.py
+++ b/tests/test_session_errors.py
@@ -13,8 +13,8 @@
     Client,
 )
 from lmstudio.async_api import (
-    AsyncSession,
-    AsyncSessionSystem,
+    _AsyncSession,
+    _AsyncSessionSystem,
 )
 from lmstudio.sync_api import (
     SyncLMStudioWebsocket,
@@ -34,7 +34,7 @@
 from .support.lmstudio import ErrFunc
 
 
-async def check_call_errors_async(session: AsyncSession) -> None:
+async def check_call_errors_async(session: _AsyncSession) -> None:
     # Remote calls on the underlying websocket are expected to fail when not connected
     with pytest.raises(
         LMStudioWebsocketError,
@@ -57,7 +57,7 @@ async def check_call_errors_async(session: AsyncSession) -> None:
 @pytest.mark.asyncio
 async def test_session_not_started_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
-    session = AsyncSessionSystem(AsyncClient())
+    session = _AsyncSessionSystem(AsyncClient())
     # Sessions start out disconnected
     assert not session.connected
     # Check server call errors are reported as expected
@@ -69,7 +69,7 @@ async def test_session_not_started_async(caplog: LogCap) -> None:
 async def test_session_disconnected_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
     client = AsyncClient()
-    session = AsyncSessionSystem(client)
+    session = _AsyncSessionSystem(client)
     async with client._task_manager, session:
         assert session.connected
     # Session is disconnected after use
@@ -82,7 +82,7 @@ async def test_session_disconnected_async(caplog: LogCap) -> None:
 async def test_session_closed_port_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
     client = AsyncClient(closed_api_host())
-    session = AsyncSessionSystem(client)
+    session = _AsyncSessionSystem(client)
     # Sessions start out disconnected
     assert not session.connected
     # Should get an SDK exception rather than the underlying exception
@@ -101,7 +101,7 @@ async def test_session_nonresponsive_port_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
     with nonresponsive_api_host() as api_host:
         client = AsyncClient(api_host)
-        session = AsyncSessionSystem(client)
+        session = _AsyncSessionSystem(client)
         # Sessions start out disconnected
         assert not session.connected
         # Should get an SDK exception rather than the underlying exception
diff --git a/tests/test_sessions.py b/tests/test_sessions.py
index 9dce3c1..e93c87c 100644
--- a/tests/test_sessions.py
+++ b/tests/test_sessions.py
@@ -12,9 +12,9 @@
     LMStudioWebsocketError,
 )
 from lmstudio.async_api import (
-    AsyncLMStudioWebsocket,
-    AsyncSession,
-    AsyncSessionSystem,
+    _AsyncLMStudioWebsocket,
+    _AsyncSession,
+    _AsyncSessionSystem,
 )
 from lmstudio.sync_api import (
     SyncLMStudioWebsocket,
@@ -27,7 +27,7 @@
 from .support import LOCAL_API_HOST
 
 
-async def check_connected_async_session(session: AsyncSession) -> None:
+async def check_connected_async_session(session: _AsyncSession) -> None:
     assert session.connected
     session_ws = session._lmsws
     assert session_ws is not None
@@ -50,7 +50,7 @@ async def check_connected_async_session(session: AsyncSession) -> None:
 async def test_session_cm_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
     client = AsyncClient()
-    session = AsyncSessionSystem(client)
+    session = _AsyncSessionSystem(client)
     # Sessions start out disconnected
     assert not session.connected
     # Disconnecting should run without error
@@ -156,7 +156,7 @@ async def test_websocket_cm_async(caplog: LogCap) -> None:
     caplog.set_level(logging.DEBUG)
     auth_details = AsyncClient._format_auth_message()
     tm = AsyncTaskManager(on_activation=None)
-    lmsws = AsyncLMStudioWebsocket(tm, f"http://{LOCAL_API_HOST}/system", auth_details)
+    lmsws = _AsyncLMStudioWebsocket(tm, f"http://{LOCAL_API_HOST}/system", auth_details)
     # SDK client websockets start out disconnected
     assert not lmsws.connected
     # Entering the CM opens the websocket if it isn't already open
diff --git a/tests/unload_models.py b/tests/unload_models.py
index 675651e..d8fcdfd 100644
--- a/tests/unload_models.py
+++ b/tests/unload_models.py
@@ -10,7 +10,9 @@
     TOOL_LLM_ID,
 )
 
-AsyncSessionModel = lms.async_api.AsyncSessionEmbedding | lms.async_api.AsyncSessionLlm
+AsyncSessionModel = (
+    lms.async_api._AsyncSessionEmbedding | lms.async_api._AsyncSessionLlm
+)
 
 
 async def _unload_model(session: AsyncSessionModel, model_identifier: str) -> None:
diff --git a/tox.ini b/tox.ini
index 642c9cb..d605fba 100644
--- a/tox.ini
+++ b/tox.ini
@@ -25,11 +25,11 @@ commands =
 
 [testenv:load-test-models]
 commands =
-    python -W "ignore:Note the async API is not yet stable:FutureWarning" -m tests.load_models
+    python -m tests.load_models
 
 [testenv:unload-test-models]
 commands =
-    python -W "ignore:Note the async API is not yet stable:FutureWarning" -m tests.unload_models
+    python -m tests.unload_models
 
 [testenv:coverage]
 # Subprocess coverage based on https://hynek.me/articles/turbo-charge-tox/