huggingface/smolagents · error · ValueError

Received both `token` and `api_key` arguments. Please provid

Error message

Received both `token` and `api_key` arguments. Please provide only one of them. `api_key` is an alias for `token` to make the API compatible with OpenAI's client. It has the exact same behavior as `token`.

What it means

InferenceClientModel.__init__ accepts both `token` and `api_key` (an OpenAI-compatibility alias for token), but passing both is ambiguous so it raises ValueError immediately. Only one credential argument may be supplied; they are otherwise identical in behavior.

Source

Thrown at src/smolagents/models.py:1528

    "Quantum mechanics is the branch of physics that studies..."
    ```
    """

    def __init__(
        self,
        model_id: str = "Qwen/Qwen3-Next-80B-A3B-Thinking",
        provider: str | None = None,
        token: str | None = None,
        timeout: int = 120,
        client_kwargs: dict[str, Any] | None = None,
        custom_role_conversions: dict[str, str] | None = None,
        api_key: str | None = None,
        bill_to: str | None = None,
        base_url: str | None = None,
        **kwargs,
    ):
        if token is not None and api_key is not None:
            raise ValueError(
                "Received both `token` and `api_key` arguments. Please provide only one of them."
                " `api_key` is an alias for `token` to make the API compatible with OpenAI's client."
                " It has the exact same behavior as `token`."
            )
        token = token if token is not None else api_key
        if token is None:
            token = os.getenv("HF_TOKEN")
        self.client_kwargs = {
            **(client_kwargs or {}),
            "model": model_id,
            "provider": provider,
            "token": token,
            "timeout": timeout,
            "bill_to": bill_to,
            "base_url": base_url,
        }
        super().__init__(model_id=model_id, custom_role_conversions=custom_role_conversions, **kwargs)

View on GitHub (pinned to 30bb116109)

Solutions

1. Pass only one of the two: keep `api_key=` for OpenAI-style code, or `token=` — never both.
2. If a wrapper injects token, remove it from your explicit call (or filter kwargs) so only one credential reaches the constructor.

Example fix

# before
model = InferenceClientModel(model_id="...", token=os.environ["HF_TOKEN"], api_key=os.environ["HF_TOKEN"])

# after
model = InferenceClientModel(model_id="...", api_key=os.environ["HF_TOKEN"])

creds = {k: v for k, v in {"token": token, "api_key": api_key}.items() if v is not None}
assert len(creds) <= 1, "pass only one of token/api_key"

null

try:
    model = InferenceClientModel(model_id=mid, api_key=key)
except ValueError as e:
    if "both `token` and `api_key`" in str(e):
        model = InferenceClientModel(model_id=mid, api_key=key)  # retry with single credential
    else:
        raise

When it happens

Trigger: Constructing InferenceClientModel(token=..., api_key=...) in the same call — commonly when code sets api_key explicitly while token is also filled from a config/env wrapper or kwargs merging.

Common situations: Migrating OpenAI-style code to smolagents and passing api_key while a framework (e.g. an agent scaffold) injects token; copying examples that use token and adding api_key on top; forwarding **kwargs that happen to contain token.

Related errors

• InferenceClientModel only supports structured outputs with t

AI-assisted analysis of huggingface/smolagents@30bb116109 (2026-08-28). Data as JSON: /api/errors/6746480e44ebd98d. Report an issue: GitHub.