xtekky/gpt4free · error · ModelNotFoundError

Model '{_model}' is not supported by LMArena provider.

Error message

Model '{_model}' is not supported by LMArena provider.

What it means

Raised by the get_mode_id() closure in LMArena.create_async_generator. After applying model_aliases, the requested model name was not found in text_models, image_models, or video_models, so it cannot be mapped to an LMArena model ID. ModelNotFoundError tells the caller the model string is invalid for this provider.

Source

Thrown at g4f/Provider/needs_auth/LMArena.py:594

            if not cls._models_loaded:
                # change to async
                await cls.get_models_async()

            def get_mode_id(_model):
                model_id = None
                # if not model:
                #     model = cls.default_model
                if _model in cls.model_aliases:
                    _model = cls.model_aliases[_model]
                if _model in cls.text_models:
                    model_id = cls.text_models[_model]
                elif _model in cls.image_models:
                    model_id = cls.image_models[_model]
                elif _model in cls.video_models:
                    model_id = cls.video_models[_model]
                elif _model:
                    raise ModelNotFoundError(
                        f"Model '{_model}' is not supported by LMArena provider."
                    )
                return model_id

            modelA: str = model
            modelB: str = kwargs.get("modelB", "")
            modelAId = get_mode_id(modelA)
            modelBId = get_mode_id(modelB) if modelB else None
            if modelAId and modelBId:
                mode = "side-by-side"
            elif modelAId:
                mode = "direct-battle"
            else:
                mode = "battle"
            if conversation and getattr(conversation, "evaluationSessionId", None):
                url = cls.post_to_evaluation.format(id=conversation.evaluationSessionId)
                evaluationSessionId = conversation.evaluationSessionId
            else:

View on GitHub (pinned to 973504e177)

Solutions

  1. List supported models first and pick from them: models = await LMArena.get_models_async(); use a key from that result
  2. Update g4f — model aliases and ID maps for LMArena change frequently
  3. If the model list is empty, ensure auth worked (models load only after a valid session) and retry
  4. Check model_aliases for the provider to use the accepted alias spelling

Example fix

# before
resp = await client.chat.completions.create(model='gpt-4-turbo', messages=msgs)  # not an LMArena model

# after
models = await LMArena.get_models_async()  # or check g4f model listing
resp = await client.chat.completions.create(model=list(models)[0], messages=msgs)
Defensive patterns

Strategy: validation

Validate before calling

models = await LMArena.get_models_async()
if model not in models and model not in LMArena.model_aliases:
    model = next(iter(models))  # or raise a clear config error

Type guard

def model_supported(name: str, models: dict, aliases: dict) -> bool:
    return name in aliases or name in models

Try / catch

from g4f.errors import ModelNotFoundError
try:
    resp = await client.chat.completions.create(model=model, messages=msgs)
except ModelNotFoundError:
    models = await LMArena.get_models_async()
    resp = await client.chat.completions.create(model=list(models)[0], messages=msgs)

Prevention

When it happens

Trigger: Passing a model string that is not in LMArena.get_models() output, e.g. a model name belonging to another provider, a renamed/removed arena model, or a typo; also modelB resolution when kwargs['modelB'] names an unknown model.

Common situations: Copying model names from other providers ('gpt-4o' style names that LMArena does not expose), using a stale hardcoded model list after LMArena rotated its arena models, or the models list failing to load (so dicts are empty and every name misses).

Related errors

AI-assisted analysis of xtekky/gpt4free@973504e177 (2026-08-14). Data as JSON: /api/errors/28b40d65c18b34f0. Report an issue: GitHub.