BerriAI/litellm · error · Exception

Model not found in cost map for model=

Error message

Model not found in cost map for model={model}

What it means

default_video_cost_calculator (litellm/cost_calculator.py:2026) prices video generation by looking up the model (with and without its provider prefix) in litellm.model_cost, or by using the deployment-level model_info when supplied. If no cost entry is found under any probed key it raises this Exception. Note the asymmetry: if an entry IS found but lacks per-second pricing, the function logs and returns 0.0 — only a completely missing entry raises.

Solutions

1. Register video pricing before the call: litellm.register_model({"<model>": {"output_cost_per_video_per_second": 0.4, "mode": "video_generation"}}).
2. On LiteLLM proxy, set custom pricing on the deployment's model_info (output_cost_per_video_per_second, or tiered output_cost_per_second_per_resolution) — default_video_cost_calculator consumes model_info first.
3. Upgrade litellm so the bundled model_prices_and_context_window.json covers the new video model.
4. Check the exact model string: the lookup tries model, model.split("/")[-1], and "<custom_llm_provider>/<model>" — make your registered key match one of them.

Example fix

# before: no entry for the video model -> raise on cost calculation
litellm.register_model({"veo-3-fast": {"mode": "video_generation"}})

# after: per-second pricing key present
litellm.register_model({
    "gemini/veo-3-fast": {
        "output_cost_per_video_per_second": 0.15,
        "mode": "video_generation",
    }
})

import litellm

def ensure_video_pricing(model: str, custom_llm_provider: str | None = None) -> None:
    candidates = [model, model.split("/")[-1]]
    if custom_llm_provider:
        candidates.append(f"{custom_llm_provider}/{model}")
    if not any(c in litellm.model_cost for c in candidates):
        raise ValueError(
            f"No video pricing for {model!r}; register output_cost_per_video_per_second first."
        )

When it happens

Trigger: Calling a video-generation endpoint (e.g. litellm proxy /v1/videos or a vertex/gemini veo model) whose model string has no key in litellm.model_cost and whose deployment carries no model_info pricing: new veo/sora-style models on an older litellm, custom video model names, or a provider-prefixed string that matches neither "<provider>/<model>" nor the bare model.

Common situations: New video models shipping faster than the pinned litellm pricing file; proxy deployments that forgot to set custom pricing on the video deployment; model strings passed with unusual casing or aliasing so none of base_model_name / model / model_without_provider / prefixed_model match a key.

Related errors

• Model not found in cost map. Tried checking
• No pricing information found for model
• Could not calculate cost for model
• Cannot delete file . The file is referenced by batch(es) in…
• Discount for must be a number

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/bb2131cc3b4607e2. Report an issue: GitHub.

Appendix: source

Thrown at litellm/cost_calculator.py:2086

        models_to_check: Final[list[str | None]] = [
            base_model_name,
            model,
            model_without_provider,
            model_name_without_custom_llm_provider,
        ]
        for _model in models_to_check:
            if _model is not None and _model in litellm.model_cost:
                cost_info = litellm.model_cost[_model]
                break

        # If still not found, try with custom_llm_provider prefix
        if cost_info is None and custom_llm_provider:
            prefixed_model: Final = f"{custom_llm_provider}/{model}"
            if prefixed_model in litellm.model_cost:
                cost_info = litellm.model_cost[prefixed_model]

    if cost_info is None:
        raise Exception(f"Model not found in cost map for model={model}")

    # Check for video-specific cost per second first
    video_cost_per_second: Final = cost_info.get("output_cost_per_video_per_second")
    if video_cost_per_second is not None:
        return video_cost_per_second * duration_seconds

    output_cost_per_second: Final = _video_output_cost_per_second(cost_info, video_resolution)
    if output_cost_per_second is not None:
        return output_cost_per_second * duration_seconds

    # If no cost information found, return 0
    verbose_logger.info(
        "No cost information found for video model %s. Please add pricing to model_prices_and_context_window.json",
        model,
    )
    return 0.0

View on GitHub (pinned to 77b7c6c40c)