livekit · demianarc · Aug 13, 2026
@@ -860,9 +860,9 @@ def with_telnyx(
     @staticmethod
     def with_nebius(
         *,
-        model: str | NebiusChatModels = "meta-llama/Meta-Llama-3.1-70B-Instruct",
+        model: str | NebiusChatModels,
         api_key: str | None = None,
-        base_url: str = "https://api.studio.nebius.com/v1/",
+        base_url: str = "https://api.tokenfactory.nebius.com/v1/",
         client: openai.AsyncClient | None = None,
         user: NotGivenOr[str] = NOT_GIVEN,
         temperature: NotGivenOr[float] = NOT_GIVEN,
@@ -874,10 +874,11 @@ def with_nebius(
         top_p: NotGivenOr[float] = NOT_GIVEN,
     ) -> LLM:
         """
-        Create a new instance of Nebius LLM.
+        Create a new Nebius Token Factory LLM.
 
-        ``api_key`` must be set to your Nebius API key, either using the argument or by setting
-        the ``NEBIUS_API_KEY`` environmental variable.
+        ``model`` must be a current chat-completions model from the Token Factory model catalog.
+        ``api_key`` must be set to your Token Factory API key, either using the argument or by
+        setting the ``NEBIUS_API_KEY`` environmental variable.
         """
 
         api_key = api_key or os.environ.get("NEBIUS_API_KEY")

@@ -80,33 +80,19 @@
 ]
 
 NebiusChatModels = Literal[
-    "meta-llama/Meta-Llama-3.1-70B-Instruct",
     "meta-llama/Llama-3.3-70B-Instruct",
-    "meta-llama/Llama-3.3-8B-Instruct",
-    "meta-llama/Meta-Llama-3.1-405B-Instruct",
     "openai/gpt-oss-120b",
-    "openai/gpt-oss-20b",
-    "moonshotai/Kimi-K2-Instruct",
-    "Qwen/Qwen3-Coder-480B-A35B-Instruct",
-    "NousResearch/Hermes-4-405B",
-    "NousResearch/Hermes-4-70B",
-    "zai-org/GLM-4.5",
-    "zai-org/GLM-4.5-Air",
-    "deepseek-ai/DeepSeek-R1-0528",
-    "deepseek-ai/DeepSeek-R1",
-    "deepseek-ai/DeepSeek-V3",
-    "deepseek-ai/DeepSeek-V3-0324",
-    "Qwen/Qwen3-235B-A22B-Instruct-2507",
-    "Qwen/Qwen3-235B-A22B",
-    "Qwen/Qwen3-32B",
-    "Qwen/Qwen3-30B-A3B",
-    "Qwen/Qwen3-4B-fast",
-    "Qwen/Qwen3-14B",
-    "Qwen/Qwen2.5-Coder-7B",
-    "Qwen/Qwen2.5-Coder-32B-Instruct",
-    "nvidia/Llama-3_1-Nemotron-Ultra-253B-v1",
-    "mistralai/Mistral-Nemo-Instruct-2407",
-    "google/gemma-2-2b-it",
+    "moonshotai/Kimi-K3",
+    "moonshotai/Kimi-K2.7-Code",
+    "moonshotai/Kimi-K2.6",
+    "zai-org/GLM-5.2",
+    "zai-org/GLM-5.1",
+    "deepseek-ai/DeepSeek-V4-Pro",
+    "deepseek-ai/DeepSeek-V4-Flash",
+    "Qwen/Qwen3.5-397B-A17B",
+    "nvidia/Nemotron-3_5-Lightning",
+    "nvidia/Nemotron-3-Ultra-550b-a55b",
+    "MiniMaxAI/MiniMax-M3",
 ]
 
 CerebrasChatModels = Literal[

@@ -0,0 +1,84 @@
+from __future__ import annotations
+
+import json
+from unittest.mock import patch
+
+import httpx
+import openai
+import pytest
+
+from livekit.agents.llm import ChatContext
+from livekit.plugins.openai import LLM
+
+pytestmark = pytest.mark.unit
+
+_MODEL = "meta-llama/Llama-3.3-70B-Instruct"
+_BASE_URL = "https://api.tokenfactory.nebius.com/v1/"
+_STREAM_RESPONSE = b"""data: {"id":"chatcmpl-test","choices":[{"delta":{"content":"ok","role":"assistant"},"finish_reason":null,"index":0}],"created":0,"model":"meta-llama/Llama-3.3-70B-Instruct","object":"chat.completion.chunk"}
+
+data: {"id":"chatcmpl-test","choices":[{"delta":{},"finish_reason":"stop","index":0}],"created":0,"model":"meta-llama/Llama-3.3-70B-Instruct","object":"chat.completion.chunk"}
+
+data: [DONE]
+
+"""
+
+
+class _RecordingTransport(httpx.AsyncBaseTransport):
+    def __init__(self) -> None:
+        self.requests: list[httpx.Request] = []
+
+    async def handle_async_request(self, request: httpx.Request) -> httpx.Response:
+        await request.aread()
+        self.requests.append(request)
+        return httpx.Response(
+            200,
+            headers={"content-type": "text/event-stream"},
+            content=_STREAM_RESPONSE,
+            request=request,
+        )
+
+
+def test_nebius_helper_configures_token_factory(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("NEBIUS_API_KEY", "env-key")
+
+    with patch("livekit.plugins.openai.llm.openai.AsyncClient") as client_cls:
+        client = client_cls.return_value
+        llm = LLM.with_nebius(model=_MODEL)
+
+    assert llm.model == _MODEL
+    assert client_cls.call_args.kwargs["api_key"] == "env-key"
+    assert str(client_cls.call_args.kwargs["base_url"]) == _BASE_URL
+    assert client is llm._client
+
+
+async def test_nebius_helper_sends_request_to_configured_client() -> None:
+    transport = _RecordingTransport()
+    llm = LLM.with_nebius(
+        model=_MODEL,
+        api_key="test-key",
+        base_url=_BASE_URL,
+        client=None,
+    )
+    await llm._client.close()
+    llm._client = openai.AsyncClient(
+        api_key="test-key",
+        base_url=_BASE_URL,
+        http_client=httpx.AsyncClient(transport=transport),
+    )
+    chat_ctx = ChatContext.empty()
+    chat_ctx.add_message(role="user", content="hello")
+
+    stream = llm.chat(chat_ctx=chat_ctx)
+    try:
+        async for _ in stream:
+            pass
+    finally:
+        await stream.aclose()
+        await llm.aclose()
+
+    assert len(transport.requests) == 1
+    request = transport.requests[0]
+    assert str(request.url) == f"{_BASE_URL}chat/completions"
+    payload = json.loads(request.content)
+    assert payload["model"] == _MODEL
+    assert payload["messages"] == [{"role": "user", "content": "hello"}]