invoke-ai/InvokeAI · error · ValueError

Unsupported Z-Image model format: {transformer_config.format

Error message

Unsupported Z-Image model format: {transformer_config.format}

What it means

Z-Image only knows how to load its transformer for Diffusers/Checkpoint (unquantized) and GGUF/SDNQ (quantized) model formats. Any other ModelFormat on the transformer config has no supported loading/patching path, so _run_diffusion raises ValueError with the offending format.

Source

Thrown at invokeai/app/invocations/z_image_denoise.py:444

            # For Heun scheduler, the number of actual steps may differ
            num_scheduler_steps = len(scheduler.timesteps)
        else:
            num_scheduler_steps = total_steps

        with ExitStack() as exit_stack:
            # Get transformer config to determine if it's quantized
            transformer_config = context.models.get_config(self.transformer.transformer)

            # Determine if the model is quantized.
            # If the model is quantized, then we need to apply the LoRA weights as sidecar layers. This results in
            # slower inference than direct patching, but is agnostic to the quantization format.
            if transformer_config.format in [ModelFormat.Diffusers, ModelFormat.Checkpoint]:
                model_is_quantized = False
            elif transformer_config.format in [ModelFormat.GGUFQuantized, ModelFormat.SDNQQuantized]:
                model_is_quantized = True
            else:
                raise ValueError(f"Unsupported Z-Image model format: {transformer_config.format}")

            # Load transformer - always use base transformer, control is handled via extension
            (cached_weights, transformer) = exit_stack.enter_context(transformer_info.model_on_device())

            # Prepare control extension if control is provided
            control_extension: ZImageControlNetExtension | None = None

            if self.control is not None:
                # Load control adapter using context manager (proper GPU memory management)
                control_model_info = context.models.load(self.control.control_model)
                (_, control_adapter) = exit_stack.enter_context(control_model_info.model_on_device())
                assert isinstance(control_adapter, ZImageControlAdapter)

                # Get control_in_dim from adapter config (16 for V1, 33 for V2.0)
                adapter_config = control_adapter.config
                control_in_dim = adapter_config.get("control_in_dim", 16)
                num_control_blocks = adapter_config.get("num_control_blocks", 6)

View on GitHub (pinned to 0b6a024f2f)

Solutions

1. Install/convert the Z-Image model to Diffusers, Checkpoint, GGUF, or SDNQ format
2. Fix the model's format field in the model manager config to match the actual on-disk format
3. Re-download the model from a source providing a supported format
4. Check InvokeAI version/update notes for newly supported Z-Image formats

Example fix

// before
# model config: format = 'BnbQuantized'
transformer_info = context.models.get_by_key(z_image_model_key)
// after
# convert/reconfigure model so format in {Diffusers, Checkpoint, GGUFQuantized, SDNQQuantized}
transformer_info = context.models.get_by_key(z_image_model_key)

from invokeai.backend.model_manager.config import ModelFormat
SUPPORTED = {ModelFormat.Diffusers, ModelFormat.Checkpoint, ModelFormat.GGUFQuantized, ModelFormat.SDNQQuantized}
config = context.models.get_config(z_image_model_key)
assert config.format in SUPPORTED, f"Unsupported Z-Image format: {config.format}"

def is_supported_zimage_format(cfg) -> bool:
    return cfg.format in {ModelFormat.Diffusers, ModelFormat.Checkpoint, ModelFormat.GGUFQuantized, ModelFormat.SDNQQuantized}

try:
    output = denoise.invoke(context)
except ValueError as e:
    if "Unsupported Z-Image model format" in str(e):
        raise ModelFormatError("convert the model to Diffusers/Checkpoint/GGUF/SDNQ") from e
    raise

When it happens

Trigger: Selecting a Z-Image main model whose ModelConfig.format is not one of Diffusers, Checkpoint, GGUFQuantized, or SDNQQuantized — invoke() then reaches the format switch in _run_diffusion and falls into the else branch.

Common situations: Pointing the model manager at a model converted with an unsupported tool (e.g. bnb quantization) or an exotic/custom format; a model-manager config whose format enum was set incorrectly; using a model file from a different architecture's pipeline.

Related errors

• Unsupported model format: {config.format}
• The {model_name} model must be a Diffusers-style Z-Image pip
• The {model_name} model must be a Diffusers format model. The
• The {model_name} model must be a Diffusers-style FLUX.2 pipe
• To extract the VAE and Qwen3-VL encoder, the {model_name} mo

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/a06de20a7e0f1864. Report an issue: GitHub.