BerriAI/litellm · error · Exception

Invalid arg. Model cannot be none.

Error message

Invalid arg. Model cannot be none.

What it means

The generic cost calculator (completion_cost) requires a model name to look up per-token pricing. If model is None — neither passed explicitly nor recoverable from the completion_response — it raises this generic Exception immediately, before reconstructing the usage block.

Source

Thrown at litellm/cost_calculator.py:353

    Parameters:
        model (str): The name of the model to use. Default is ""
        prompt_tokens (int): The number of tokens in the prompt.
        completion_tokens (int): The number of tokens in the completion.
        response_time (float): The amount of time, in milliseconds, it took the call to complete.
        prompt_characters (float): The number of characters in the prompt. Used for vertex ai cost calculation.
        completion_characters (float): The number of characters in the completion response. Used for vertex ai cost calculation.
        custom_llm_provider (str): The llm provider to whom the call was made (see init.py for full list)
        custom_cost_per_token: Optional[CostPerToken]: the cost per input + output token for the llm api call.
        custom_cost_per_second: Optional[float]: the cost per second for the llm api call.
        call_type: Optional[str]: the call type

    Returns:
        tuple: A tuple containing the cost in USD dollars for prompt tokens and completion tokens, respectively.
    """

    if model is None:
        raise Exception("Invalid arg. Model cannot be none.")

    ## RECONSTRUCT USAGE BLOCK ##
    if usage_object is not None:
        usage_block = usage_object
    else:
        usage_block = Usage(
            prompt_tokens=prompt_tokens,
            completion_tokens=completion_tokens,
            total_tokens=prompt_tokens + completion_tokens,
            cache_creation_input_tokens=cache_creation_input_tokens,
            cache_read_input_tokens=cache_read_input_tokens,
        )

    ## CUSTOM PRICING ##
    # Normalize cache token counts across providers:
    #   - OpenAI-compatible: usage.prompt_tokens_details.cached_tokens
    #     (prompt_tokens already INCLUDES cached_tokens)
    #   - Anthropic: usage.cache_read_input_tokens / cache_creation_input_tokens

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Pass model='provider/model-name' explicitly to completion_cost.
2. Ensure the completion_response object includes a 'model' attribute — LiteLLM normally extracts it via completion_response.get('model').
3. In tests, construct ModelResponse(..., model='gpt-4o-mini') or copy a real response shape.
4. If wrapping LiteLLM, thread the original request's model through to the cost-calculation step.

Example fix

# before
cost = litellm.completion_cost(completion_response=mock_resp)  # model missing

# after
cost = litellm.completion_cost(
    completion_response=mock_resp,
    model="gpt-4o-mini",
    custom_llm_provider="openai",
)

if model is None:
    model = completion_response.get("model") if completion_response else None
if model is None:
    raise ValueError("model required for cost calculation")

def has_model_for_cost(model: str | None, resp) -> bool:
    return model is not None or bool(resp and resp.get("model"))

try:
    cost = litellm.completion_cost(completion_response=resp, model=model)
except Exception as e:
    if "Model cannot be none" in str(e):
        cost = 0.0  # or re-raise with request context
    else:
        raise

When it happens

Trigger: Calling litellm.completion_cost(model=None, completion_response=resp) where resp also lacks a 'model' field (custom ModelResponse, mocked responses, or stripped provider payloads); constructing ModelResponse manually and passing it to cost functions without setting .get('model').

Common situations: Unit tests with hand-built response objects; streaming code that builds a cost-call from a chunk that never carried the model; proxy handlers that drop the model field during serialization.

Related errors

• prompt_characters must be provided for tts calls. prompt_cha
• Model is None and does not exist in passed completion_respon
• soft_budget cannot be negative. Received: {data.soft_budget}
• soft_budget ({data.soft_budget}) must be strictly lower than
• Model '{m}' not in team's allowed models. Team allowed model

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/916b6da80d01ba59. Report an issue: GitHub.