invoke-ai/InvokeAI · error · Exception

Unknown model: {model_key}

Error message

Unknown model: {model_key}

What it means

The SDXL main-model loader raises this when the ModelField's key (`self.model.key`) does not exist in the model manager's record store at invoke time. The TODO comment notes proper not-found exceptions were intended; a generic Exception is used as a placeholder. It means the referenced model was deleted, never installed, or the queue item references stale model state.

Source

Thrown at invokeai/app/invocations/sdxl.py:43

@invocation("sdxl_model_loader", title="Main Model - SDXL", tags=["model", "sdxl"], category="model", version="1.0.4")
class SDXLModelLoaderInvocation(BaseInvocation):
    """Loads an sdxl base model, outputting its submodels."""

    model: ModelIdentifierField = InputField(
        description=FieldDescriptions.sdxl_main_model,
        ui_model_base=BaseModelType.StableDiffusionXL,
        ui_model_type=ModelType.Main,
    )
    # TODO: precision?

    def invoke(self, context: InvocationContext) -> SDXLModelLoaderOutput:
        model_key = self.model.key

        # TODO: not found exceptions
        if not context.models.exists(model_key):
            raise Exception(f"Unknown model: {model_key}")

        unet = self.model.model_copy(update={"submodel_type": SubModelType.UNet})
        scheduler = self.model.model_copy(update={"submodel_type": SubModelType.Scheduler})
        tokenizer = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer})
        text_encoder = self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder})
        tokenizer2 = self.model.model_copy(update={"submodel_type": SubModelType.Tokenizer2})
        text_encoder2 = self.model.model_copy(update={"submodel_type": SubModelType.TextEncoder2})
        vae = self.model.model_copy(update={"submodel_type": SubModelType.VAE})

        return SDXLModelLoaderOutput(
            unet=UNetField(unet=unet, scheduler=scheduler, loras=[]),
            clip=CLIPField(tokenizer=tokenizer, text_encoder=text_encoder, loras=[], skipped_layers=0),
            clip2=CLIPField(tokenizer=tokenizer2, text_encoder=text_encoder2, loras=[], skipped_layers=0),
            vae=VAEField(vae=vae),
        )


@invocation(

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Install the referenced SDXL model (or fix its path/scan) so the key exists, then reopen the workflow and reselect the model in the Model Loader node.
2. Re-select the model in the SDXL Main Model Loader node so the graph stores a fresh valid key.
3. If the model exists, rescan the models directory or restart so the model records rebuild.
4. Do not hand-edit workflow JSON model keys; export/import via the UI instead.

Example fix

// before
raise Exception(f"Unknown model: {model_key}")
// after
from invokeai.app.services.shared.exceptions import UnknownModelException
if not context.models.exists(model_key):
    raise UnknownModelException(f"Unknown model: {model_key}")

if not context.models.exists(model_loader.model.key):
    raise LookupError(f"Model {model_loader.model.key} is not installed")

def model_exists(context, model) -> bool:
    return context.models.exists(model.key)

try:
    output = graph_executor.invoke(context)
except Exception as e:
    if str(e).startswith("Unknown model:"):
        # re-select models in the graph and requeue
        requeue_with_valid_models(graph)
    else:
        raise

When it happens

Trigger: context.models.exists(model_key) returns False because the model record was deleted between enqueue and execution, the model was never installed/converted, a workflow JSON was shared across installs with different model IDs, or the main-model invocation's `model` field points at a nonexistent/invalid key.

Common situations: Running old saved graphs after clearing the models directory; switching between SQLite databases; renaming/moving models outside the app; importing workflows from other users; partial model downloads failing to register.

Related errors

• Admin privileges required
• Expected LlavaOnevisionForConditionalGeneration, got {type(m
• Expected ModelPatchRaw for LoRA '{lora.lora.key}', got {type
• Expected AutoencoderKLWan or FluxAutoEncoder for Anima VAE,
• Expected PreTrainedModel for text encoder, got {type(text_en

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/122ea0830500d809. Report an issue: GitHub.