feat(z-image): add Z-Image Base (undistilled) model variant support by Pfannkuchensack · Pull Request #8799 · invoke-ai/InvokeAI

Pfannkuchensack · 2026-01-28T20:42:50Z

Summary

Add ZImageVariantType enum with 'turbo' and 'zbase' variants
Auto-detect variant on import via scheduler_config.json shift value (3.0=turbo, 6.0=zbase)
Add database migration to populate variant field for existing Z-Image models
Re-add LCM scheduler with variant-aware filtering (LCM hidden for zbase)
Auto-reset scheduler to Euler when switching to zbase model if LCM selected
Update frontend to show/hide LCM option based on model variant
Add toast notification when scheduler is auto-reset
Filter LoRAs by variant compatibility in LoRA picker (LoRAs without variant are always shown, variant-specific LoRAs only shown when matching main model variant)
Add runtime variant mismatch warnings in Z-Image LoRA loaders

Z-Image Base models are undistilled and require more steps (28-50) with higher guidance (3.0-5.0), while Z-Image Turbo is distilled for ~8 steps with CFG 1.0. LCM scheduler only works with distilled (Turbo) models.

Only Tested with Diffusers version of the Model.

Related Issues / Discussions

Adds full support for Z-Image Base (undistilled) models alongside the existing Z-Image Turbo support.

QA Instructions

Import a Z-Image Turbo model - Verify variant is detected as "turbo"
Import a Z-Image Base model - Verify variant is detected as "zbase" (shift value 6.0 in scheduler_config.json)
Select Z-Image Turbo model - Verify LCM scheduler option is available
Select Z-Image Base model - Verify LCM scheduler option is hidden (only Euler and Heun shown)
With LCM selected, switch from Turbo to Base - Verify scheduler auto-resets to Euler with toast notification
Edit model in Model Manager - Verify variant field is editable for Z-Image models
LoRA variant filtering - With a Z-Image Turbo model selected, only Turbo-specific and variant-less LoRAs should appear in the LoRA picker. ZBase-specific LoRAs should be hidden, and vice versa.
LoRA variant mismatch warning - If a LoRA with mismatched variant is applied (e.g. via workflow), a warning should appear in the log

Merge Plan

Database migration (migration_26) adds variant field to existing Z-Image models. Safe to merge anytime.

Checklist

The PR has a short but descriptive title, suitable for a changelog
Tests added / updated (if applicable)
Changes to a redux slice have a corresponding migration
Documentation added / updated (if applicable)
Updated What's New copy (if doing a release after this PR)

- Add ZImageVariantType enum with 'turbo' and 'zbase' variants - Auto-detect variant on import via scheduler_config.json shift value (3.0=turbo, 6.0=zbase) - Add database migration to populate variant field for existing Z-Image models - Re-add LCM scheduler with variant-aware filtering (LCM hidden for zbase) - Auto-reset scheduler to Euler when switching to zbase model if LCM selected - Update frontend to show/hide LCM option based on model variant - Add toast notification when scheduler is auto-reset Z-Image Base models are undistilled and require more steps (28-50) with higher guidance (3.0-5.0), while Z-Image Turbo is distilled for ~8 steps with CFG 1.0. LCM scheduler only works with distilled (Turbo) models.

lstein · 2026-01-31T23:50:49Z

Issues involving Z-Image Turbo support

I found a couple of problems with support for Z-Image Turbo.

Problem 1

After pulling the PR and restarting the backend, I'm getting config validation errors on previously-installed Z-Image Turbo models, and they vanish from the model manager.

Here is the type of error message that appears during startup:

[2026-01-31 18:39:59,772]::[InvokeAI]::WARNING --> Skipping invalid model config in the database with name Z-Image Turbo (quantized). Ignoring this model. ({"key": "b640a0c8-f687-4d07-afed-c329b3dcc636", "hash": "blake3:...)
[2026-01-31 18:39:59,772]::[InvokeAI]::WARNING --> Validation error: 1 validation error for tagged-union[Main_Diffusers_SD1_Config,Main_Diffusers_SD2_Config,Main_Diffusers_SDXL_Config,Main_Diffusers_SDXLRefiner_Config,Main_Diffusers_SD3_Config,Main_Diffusers_FLUX_Config,Main_Diffusers_Flux2_Config,Main_Diffusers_CogView4_Config,Main_Diffusers_ZImage_Config,Main_Checkpoint_SD1_Config,Main_Checkpoint_SD2_Config,Main_Checkpoint_SDXL_Config,Main_Checkpoint_SDXLRefiner_Config,Main_Checkpoint_Flux2_Config,Main_Checkpoint_FLUX_Config,Main_Checkpoint_ZImage_Config,Main_BnBNF4_FLUX_Config,Main_GGUF_Flux2_Config,Main_GGUF_FLUX_Config,Main_GGUF_ZImage_Config,VAE_Checkpoint_SD1_Config,VAE_Checkpoint_SD2_Config,VAE_Checkpoint_SDXL_Config,VAE_Checkpoint_FLUX_Config,VAE_Checkpoint_Flux2_Config,VAE_Diffusers_SD1_Config,VAE_Diffusers_SDXL_Config,VAE_Diffusers_Flux2_Config,ControlNet_Checkpoint_SD1_Config,ControlNet_Checkpoint_SD2_Config,ControlNet_Checkpoint_SDXL_Config,ControlNet_Checkpoint_FLUX_Config,ControlNet_Checkpoint_ZImage_Config,ControlNet_Diffusers_SD1_Config,ControlNet_Diffusers_SD2_Config,ControlNet_Diffusers_SDXL_Config,ControlNet_Diffusers_FLUX_Config,LoRA_LyCORIS_SD1_Config,LoRA_LyCORIS_SD2_Config,LoRA_LyCORIS_SDXL_Config,LoRA_LyCORIS_FLUX_Config,LoRA_LyCORIS_ZImage_Config,LoRA_OMI_SDXL_Config,LoRA_OMI_FLUX_Config,LoRA_Diffusers_SD1_Config,LoRA_Diffusers_SD2_Config,LoRA_Diffusers_SDXL_Config,LoRA_Diffusers_FLUX_Config,LoRA_Diffusers_ZImage_Config,ControlLoRA_LyCORIS_FLUX_Config,T5Encoder_T5Encoder_Config,T5Encoder_BnBLLMint8_Config,Qwen3Encoder_Qwen3Encoder_Config,Qwen3Encoder_Checkpoint_Config,Qwen3Encoder_GGUF_Config,TI_File_SD1_Config,TI_File_SD2_Config,TI_File_SDXL_Config,TI_Folder_SD1_Config,TI_Folder_SD2_Config,TI_Folder_SDXL_Config,IPAdapter_InvokeAI_SD1_Config,IPAdapter_InvokeAI_SD2_Config,IPAdapter_InvokeAI_SDXL_Config,IPAdapter_Checkpoint_SD1_Config,IPAdapter_Checkpoint_SD2_Config,IPAdapter_Checkpoint_SDXL_Config,IPAdapter_Checkpoint_FLUX_Config,T2IAdapter_Diffusers_SD1_Config,T2IAdapter_Diffusers_SDXL_Config,Spandrel_Checkpoint_Config,CLIPEmbed_Diffusers_G_Config,CLIPEmbed_Diffusers_L_Config,CLIPVision_Diffusers_Config,SigLIP_Diffusers_Config,FLUXRedux_Checkpoint_Config,LlavaOnevision_Diffusers_Config,Unknown_Config]

  Input tag 'main.gguf_quantized.z-image' found using get_model_discriminator_value() does not match any of the expected tags: 'main.diffusers.sd-1', 'main.diffusers.sd-2', 'main.diffusers.sdxl', 'main.diffusers.sdxl-refiner', 'main.diffusers.sd-3', 'main.diffusers.flux', 'main.diffusers.flux2', 'main.diffusers.cogview4', 'main.diffusers.z-image', 'main.checkpoint.sd-1', 'main.checkpoint.sd-2', 'main.checkpoint.sdxl', 'main.checkpoint.sdxl-refiner', 'main.checkpoint.flux2', 'main.checkpoint.flux', 'main.checkpoint.z-image.turbo', 'main.bnb_quantized_nf4b.flux', 'main.gguf_quantized.flux2', 'main.gguf_quantized.flux', 'main.gguf_quantized.z-image.turbo', 'vae.checkpoint.sd-1', 'vae.checkpoint.sd-2', 'vae.checkpoint.sdxl', 'vae.checkpoint.flux', 'vae.checkpoint.flux2', 'vae.diffusers.sd-1', 'vae.diffusers.sdxl', 'vae.diffusers.flux2', 'controlnet.checkpoint.sd-1', 'controlnet.checkpoint.sd-2', 'controlnet.checkpoint.sdxl', 'controlnet.checkpoint.flux', 'controlnet.checkpoint.z-image', 'controlnet.diffusers.sd-1', 'controlnet.diffusers.sd-2', 'controlnet.diffusers.sdxl', 'controlnet.diffusers.flux', 'lora.lycoris.sd-1', 'lora.lycoris.sd-2', 'lora.lycoris.sdxl', 'lora.lycoris.flux', 'lora.lycoris.z-image', 'lora.omi.sdxl', 'lora.omi.flux', 'lora.diffusers.sd-1', 'lora.diffusers.sd-2', 'lora.diffusers.sdxl', 'lora.diffusers.flux', 'lora.diffusers.z-image', 'control_lora.lycoris.flux', 't5_encoder.t5_encoder.any', 't5_encoder.bnb_quantized_int8b.any', 'qwen3_encoder.qwen3_encoder.any', 'qwen3_encoder.checkpoint.any', 'qwen3_encoder.gguf_quantized.any', 'embedding.embedding_file.sd-1', 'embedding.embedding_file.sd-2', 'embedding.embedding_file.sdxl', 'embedding.embedding_folder.sd-1', 'embedding.embedding_folder.sd-2', 'embedding.embedding_folder.sdxl', 'ip_adapter.invokeai.sd-1', 'ip_adapter.invokeai.sd-2', 'ip_adapter.invokeai.sdxl', 'ip_adapter.checkpoint.sd-1', 'ip_adapter.checkpoint.sd-2', 'ip_adapter.checkpoint.sdxl', 'ip_adapter.checkpoint.flux', 't2i_adapter.diffusers.sd-1', 't2i_adapter.diffusers.sdxl', 'spandrel_image_to_image.checkpoint.any', 'clip_embed.diffusers.any.gigantic', 'clip_embed.diffusers.any.large', 'clip_vision.diffusers.any', 'siglip.diffusers.any', 'flux_redux.checkpoint.flux', 'llava_onevision.diffusers.any', 'unknown.unknown.unknown' [type=union_tag_invalid, input_value={'key': 'b640a0c8-f687-4d...ed', 'variant': 'turbo'}, input_type=dict]

    For further information visit https://errors.pydantic.dev/2.12/v/union_tag_invalid

Problem 2

Installing a new ZiT model (in this case z_image_turbo-Q4_K.gguf) fails with a similar pydantic validation error:

[2026-01-31 18:46:27,782]::[ModelInstallService]::ERROR --> Model install error: https://huggingface.co/leejet/Z-Image-Turbo-GGUF/resolve/main/z_image_turbo-Q4_K.gguf
ValidationError: 1 validation error for tagged-union[Main_Diffusers_SD1_Config,Main_Diffusers_SD2_Config,Main_Diffusers_SDXL_Config,Main_Diffusers_SDXLRefiner_Config,Main_Diffusers_SD3_Config,Main_Diffusers_FLUX_Config,Main_Diffusers_Flux2_Config,Main_Diffusers_CogView4_Config,Main_Diffusers_ZImage_Config,Main_Checkpoint_SD1_Config,Main_Checkpoint_SD2_Config,Main_Checkpoint_SDXL_Config,Main_Checkpoint_SDXLRefiner_Config,Main_Checkpoint_Flux2_Config,Main_Checkpoint_FLUX_Config,Main_Checkpoint_ZImage_Config,Main_BnBNF4_FLUX_Config,Main_GGUF_Flux2_Config,Main_GGUF_FLUX_Config,Main_GGUF_ZImage_Config,VAE_Checkpoint_SD1_Config,VAE_Checkpoint_SD2_Config,VAE_Checkpoint_SDXL_Config,VAE_Checkpoint_FLUX_Config,VAE_Checkpoint_Flux2_Config,VAE_Diffusers_SD1_Config,VAE_Diffusers_SDXL_Config,VAE_Diffusers_Flux2_Config,ControlNet_Checkpoint_SD1_Config,ControlNet_Checkpoint_SD2_Config,ControlNet_Checkpoint_SDXL_Config,ControlNet_Checkpoint_FLUX_Config,ControlNet_Checkpoint_ZImage_Config,ControlNet_Diffusers_SD1_Config,ControlNet_Diffusers_SD2_Config,ControlNet_Diffusers_SDXL_Config,ControlNet_Diffusers_FLUX_Config,LoRA_LyCORIS_SD1_Config,LoRA_LyCORIS_SD2_Config,LoRA_LyCORIS_SDXL_Config,LoRA_LyCORIS_FLUX_Config,LoRA_LyCORIS_ZImage_Config,LoRA_OMI_SDXL_Config,LoRA_OMI_FLUX_Config,LoRA_Diffusers_SD1_Config,LoRA_Diffusers_SD2_Config,LoRA_Diffusers_SDXL_Config,LoRA_Diffusers_FLUX_Config,LoRA_Diffusers_ZImage_Config,ControlLoRA_LyCORIS_FLUX_Config,T5Encoder_T5Encoder_Config,T5Encoder_BnBLLMint8_Config,Qwen3Encoder_Qwen3Encoder_Config,Qwen3Encoder_Checkpoint_Config,Qwen3Encoder_GGUF_Config,TI_File_SD1_Config,TI_File_SD2_Config,TI_File_SDXL_Config,TI_Folder_SD1_Config,TI_Folder_SD2_Config,TI_Folder_SDXL_Config,IPAdapter_InvokeAI_SD1_Config,IPAdapter_InvokeAI_SD2_Config,IPAdapter_InvokeAI_SDXL_Config,IPAdapter_Checkpoint_SD1_Config,IPAdapter_Checkpoint_SD2_Config,IPAdapter_Checkpoint_SDXL_Config,IPAdapter_Checkpoint_FLUX_Config,T2IAdapter_Diffusers_SD1_Config,T2IAdapter_Diffusers_SDXL_Config,Spandrel_Checkpoint_Config,CLIPEmbed_Diffusers_G_Config,CLIPEmbed_Diffusers_L_Config,CLIPVision_Diffusers_Config,SigLIP_Diffusers_Config,FLUXRedux_Checkpoint_Config,LlavaOnevision_Diffusers_Config,Unknown_Config]
  Input tag 'main.gguf_quantized.z-image' found using get_model_discriminator_value() does not match any of the expected tags: 'main.diffusers.sd-1', 'main.diffusers.sd-2', 'main.diffusers.sdxl', 'main.diffusers.sdxl-refiner', 'main.diffusers.sd-3', 'main.diffusers.flux', 'main.diffusers.flux2', 'main.diffusers.cogview4', 'main.diffusers.z-image', 'main.checkpoint.sd-1', 'main.checkpoint.sd-2', 'main.checkpoint.sdxl', 'main.checkpoint.sdxl-refiner', 'main.checkpoint.flux2', 'main.checkpoint.flux', 'main.checkpoint.z-image.turbo', 'main.bnb_quantized_nf4b.flux', 'main.gguf_quantized.flux2', 'main.gguf_quantized.flux', 'main.gguf_quantized.z-image.turbo', 'vae.checkpoint.sd-1', 'vae.checkpoint.sd-2', 'vae.checkpoint.sdxl', 'vae.checkpoint.flux', 'vae.checkpoint.flux2', 'vae.diffusers.sd-1', 'vae.diffusers.sdxl', 'vae.diffusers.flux2', 'controlnet.checkpoint.sd-1', 'controlnet.checkpoint.sd-2', 'controlnet.checkpoint.sdxl', 'controlnet.checkpoint.flux', 'controlnet.checkpoint.z-image', 'controlnet.diffusers.sd-1', 'controlnet.diffusers.sd-2', 'controlnet.diffusers.sdxl', 'controlnet.diffusers.flux', 'lora.lycoris.sd-1', 'lora.lycoris.sd-2', 'lora.lycoris.sdxl', 'lora.lycoris.flux', 'lora.lycoris.z-image', 'lora.omi.sdxl', 'lora.omi.flux', 'lora.diffusers.sd-1', 'lora.diffusers.sd-2', 'lora.diffusers.sdxl', 'lora.diffusers.flux', 'lora.diffusers.z-image', 'control_lora.lycoris.flux', 't5_encoder.t5_encoder.any', 't5_encoder.bnb_quantized_int8b.any', 'qwen3_encoder.qwen3_encoder.any', 'qwen3_encoder.checkpoint.any', 'qwen3_encoder.gguf_quantized.any', 'embedding.embedding_file.sd-1', 'embedding.embedding_file.sd-2', 'embedding.embedding_file.sdxl', 'embedding.embedding_folder.sd-1', 'embedding.embedding_folder.sd-2', 'embedding.embedding_folder.sdxl', 'ip_adapter.invokeai.sd-1', 'ip_adapter.invokeai.sd-2', 'ip_adapter.invokeai.sdxl', 'ip_adapter.checkpoint.sd-1', 'ip_adapter.checkpoint.sd-2', 'ip_adapter.checkpoint.sdxl', 'ip_adapter.checkpoint.flux', 't2i_adapter.diffusers.sd-1', 't2i_adapter.diffusers.sdxl', 'spandrel_image_to_image.checkpoint.any', 'clip_embed.diffusers.any.gigantic', 'clip_embed.diffusers.any.large', 'clip_vision.diffusers.any', 'siglip.diffusers.any', 'flux_redux.checkpoint.flux', 'llava_onevision.diffusers.any', 'unknown.unknown.unknown' [type=union_tag_invalid, input_value={'key': 'df7115c0-ca4c-4c...ed', 'variant': 'turbo'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/union_tag_invalid

lstein · 2026-02-01T00:23:56Z

I can confirm that Z-Image Base (Diffusers):

Installs correctly
Variant is recognized correctly as "Z-Image Base"
Steps and CFG default to appropriate values
txt2img working
img2img working
Inpaint working
Outpaint working
Only Heun and Euler schedulers are shown
❌ ZiT LoRAs are working with ZiB -- Produces a fractured image
❌ Standalone denoiser .safetensors version installs -- Produces a pydantic validation error

I was unable to test whether the LCM scheduler reappears when a turbo model is installed due to the earlier issues with turbo model validation errors.

Pfannkuchensack · 2026-02-01T00:31:20Z

ZiT LoRAs are working with ZiB -- Produces a fractured image

That is to be expected

Standalone denoiser .safetensors version installs -- Produces a pydantic validation error

i will take a look

The zimage Turbo models should be getting a variant with the migration. Do you have the startup log?

lstein · 2026-02-01T03:55:17Z

The zimage Turbo models should be getting a variant with the migration. Do you have the startup log?

See Problem 1 and Problem 2 in the earlier comment for the error log startup messages. Let me know if you need more.

skunkworxdark · 2026-02-01T13:24:41Z

I reported the same issue on Discord the other day; I should have posted it here in the first place. Here is a link to the log I pasted in Discord.

https://discord.com/channels/1020123559063990373/1083864753543331981/1466506983627686062

Another example taken today using the db after updating to the latest release 6.11.0

[2026-02-01 13:11:55,731]::[InvokeAI]::WARNING --> Skipping invalid model config in the database with name Z-Image Turbo (Q8). Ignoring this model. ({"key": "7e351a8a-1201-4fd4-8b44-cb2c3672cbf7", "hash": "random:...)
[2026-02-01 13:11:55,731]::[InvokeAI]::WARNING --> Validation error: 1 validation error for tagged-union[Main_Diffusers_SD1_Config,Main_Diffusers_SD2_Config,Main_Diffusers_SDXL_Config,Main_Diffusers_SDXLRefiner_Config,Main_Diffusers_SD3_Config,Main_Diffusers_FLUX_Config,Main_Diffusers_Flux2_Config,Main_Diffusers_CogView4_Config,Main_Diffusers_ZImage_Config,Main_Checkpoint_SD1_Config,Main_Checkpoint_SD2_Config,Main_Checkpoint_SDXL_Config,Main_Checkpoint_SDXLRefiner_Config,Main_Checkpoint_Flux2_Config,Main_Checkpoint_FLUX_Config,Main_Checkpoint_ZImage_Config,Main_BnBNF4_FLUX_Config,Main_GGUF_Flux2_Config,Main_GGUF_FLUX_Config,Main_GGUF_ZImage_Config,VAE_Checkpoint_SD1_Config,VAE_Checkpoint_SD2_Config,VAE_Checkpoint_SDXL_Config,VAE_Checkpoint_FLUX_Config,VAE_Checkpoint_Flux2_Config,VAE_Diffusers_SD1_Config,VAE_Diffusers_SDXL_Config,VAE_Diffusers_Flux2_Config,ControlNet_Checkpoint_SD1_Config,ControlNet_Checkpoint_SD2_Config,ControlNet_Checkpoint_SDXL_Config,ControlNet_Checkpoint_FLUX_Config,ControlNet_Checkpoint_ZImage_Config,ControlNet_Diffusers_SD1_Config,ControlNet_Diffusers_SD2_Config,ControlNet_Diffusers_SDXL_Config,ControlNet_Diffusers_FLUX_Config,LoRA_LyCORIS_SD1_Config,LoRA_LyCORIS_SD2_Config,LoRA_LyCORIS_SDXL_Config,LoRA_LyCORIS_FLUX_Config,LoRA_LyCORIS_ZImage_Config,LoRA_OMI_SDXL_Config,LoRA_OMI_FLUX_Config,LoRA_Diffusers_SD1_Config,LoRA_Diffusers_SD2_Config,LoRA_Diffusers_SDXL_Config,LoRA_Diffusers_FLUX_Config,LoRA_Diffusers_ZImage_Config,ControlLoRA_LyCORIS_FLUX_Config,T5Encoder_T5Encoder_Config,T5Encoder_BnBLLMint8_Config,Qwen3Encoder_Qwen3Encoder_Config,Qwen3Encoder_Checkpoint_Config,Qwen3Encoder_GGUF_Config,TI_File_SD1_Config,TI_File_SD2_Config,TI_File_SDXL_Config,TI_Folder_SD1_Config,TI_Folder_SD2_Config,TI_Folder_SDXL_Config,IPAdapter_InvokeAI_SD1_Config,IPAdapter_InvokeAI_SD2_Config,IPAdapter_InvokeAI_SDXL_Config,IPAdapter_Checkpoint_SD1_Config,IPAdapter_Checkpoint_SD2_Config,IPAdapter_Checkpoint_SDXL_Config,IPAdapter_Checkpoint_FLUX_Config,T2IAdapter_Diffusers_SD1_Config,T2IAdapter_Diffusers_SDXL_Config,Spandrel_Checkpoint_Config,CLIPEmbed_Diffusers_G_Config,CLIPEmbed_Diffusers_L_Config,CLIPVision_Diffusers_Config,SigLIP_Diffusers_Config,FLUXRedux_Checkpoint_Config,LlavaOnevision_Diffusers_Config,Unknown_Config]
  Input tag 'main.gguf_quantized.z-image' found using get_model_discriminator_value() does not match any of the expected tags: 'main.diffusers.sd-1', 'main.diffusers.sd-2', 'main.diffusers.sdxl', 'main.diffusers.sdxl-refiner', 'main.diffusers.sd-3', 'main.diffusers.flux', 'main.diffusers.flux2', 'main.diffusers.cogview4', 'main.diffusers.z-image', 'main.checkpoint.sd-1', 'main.checkpoint.sd-2', 'main.checkpoint.sdxl', 'main.checkpoint.sdxl-refiner', 'main.checkpoint.flux2', 'main.checkpoint.flux', 'main.checkpoint.z-image.turbo', 'main.bnb_quantized_nf4b.flux', 'main.gguf_quantized.flux2', 'main.gguf_quantized.flux', 'main.gguf_quantized.z-image.turbo', 'vae.checkpoint.sd-1', 'vae.checkpoint.sd-2', 'vae.checkpoint.sdxl', 'vae.checkpoint.flux', 'vae.checkpoint.flux2', 'vae.diffusers.sd-1', 'vae.diffusers.sdxl', 'vae.diffusers.flux2', 'controlnet.checkpoint.sd-1', 'controlnet.checkpoint.sd-2', 'controlnet.checkpoint.sdxl', 'controlnet.checkpoint.flux', 'controlnet.checkpoint.z-image', 'controlnet.diffusers.sd-1', 'controlnet.diffusers.sd-2', 'controlnet.diffusers.sdxl', 'controlnet.diffusers.flux', 'lora.lycoris.sd-1', 'lora.lycoris.sd-2', 'lora.lycoris.sdxl', 'lora.lycoris.flux', 'lora.lycoris.z-image', 'lora.omi.sdxl', 'lora.omi.flux', 'lora.diffusers.sd-1', 'lora.diffusers.sd-2', 'lora.diffusers.sdxl', 'lora.diffusers.flux', 'lora.diffusers.z-image', 'control_lora.lycoris.flux', 't5_encoder.t5_encoder.any', 't5_encoder.bnb_quantized_int8b.any', 'qwen3_encoder.qwen3_encoder.any', 'qwen3_encoder.checkpoint.any', 'qwen3_encoder.gguf_quantized.any', 'embedding.embedding_file.sd-1', 'embedding.embedding_file.sd-2', 'embedding.embedding_file.sdxl', 'embedding.embedding_folder.sd-1', 'embedding.embedding_folder.sd-2', 'embedding.embedding_folder.sdxl', 'ip_adapter.invokeai.sd-1', 'ip_adapter.invokeai.sd-2', 'ip_adapter.invokeai.sdxl', 'ip_adapter.checkpoint.sd-1', 'ip_adapter.checkpoint.sd-2', 'ip_adapter.checkpoint.sdxl', 'ip_adapter.checkpoint.flux', 't2i_adapter.diffusers.sd-1', 't2i_adapter.diffusers.sdxl', 'spandrel_image_to_image.checkpoint.any', 'clip_embed.diffusers.any.gigantic', 'clip_embed.diffusers.any.large', 'clip_vision.diffusers.any', 'siglip.diffusers.any', 'flux_redux.checkpoint.flux', 'llava_onevision.diffusers.any', 'unknown.unknown.unknown' [type=union_tag_invalid, input_value={'key': '7e351a8a-1201-4f...ed', 'variant': 'turbo'}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.12/v/union_tag_invalid

…atch LoRA picker now hides Z-Image LoRAs with incompatible variants (e.g. ZBase LoRAs when using Turbo model). LoRAs without a variant are always shown. Backend loaders warn at runtime if a LoRA variant doesn't match the transformer variant.

Pfannkuchensack · 2026-02-13T07:03:29Z

The Errors on Startup should now be gone. Added Variant for the zimages lora. Updated the PR.

lstein

Tested and works as advertised. The official Z-Image Base is recognized and has the correct default parameters. Generation works well and the LCM scheduler is only shown when Turbo is selected.

Both Base and Turbo variant LoRAs work properly. However please confirm that the system does not try to automatically detect Turbo and Base-specific LoRA variants. The user must always manually select the variant to apply in Model edit mode. Is this correct?

Pfannkuchensack · 2026-02-17T06:42:17Z

Yes The loras are technically indistinguishable.

Pfannkuchensack added 2 commits January 28, 2026 21:38

Chore ruff format

c044e81

github-actions bot added python PRs that change python files invocations PRs that change invocations backend PRs that change backend files services PRs that change app services frontend PRs that change frontend files labels Jan 28, 2026

Chore fix windows path

b05f9b4

lstein self-assigned this Jan 28, 2026

Merge branch 'main' into feature/z-image-undistilled-support

1df28e9

lstein added the v6.12.0 Intended for 6.12.0 release label Jan 29, 2026

lstein added this to Invoke - Community Roadmap Jan 30, 2026

lstein moved this to 6.12.x in Invoke - Community Roadmap Jan 30, 2026

Merge branch 'main' into feature/z-image-undistilled-support

e592392

Pfannkuchensack added 3 commits February 8, 2026 02:40

Merge branch 'main' into feature/z-image-undistilled-support

ea15f3e

Chore typegen

e6e7c38

Pfannkuchensack marked this pull request as ready for review February 13, 2026 07:03

Pfannkuchensack requested review from JPPhoto, blessedcoolant, dunkeroni and lstein as code owners February 13, 2026 07:03

lstein approved these changes Feb 16, 2026

View reviewed changes

Pfannkuchensack and others added 2 commits February 17, 2026 14:29

Merge branch 'main' into feature/z-image-undistilled-support

de2a489

Merge branch 'main' into feature/z-image-undistilled-support

7437955

lstein enabled auto-merge (squash) February 20, 2026 00:26

lstein merged commit b0f7b55 into invoke-ai:main Feb 20, 2026
13 checks passed

Pfannkuchensack deleted the feature/z-image-undistilled-support branch February 20, 2026 01:07

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(z-image): add Z-Image Base (undistilled) model variant support#8799

feat(z-image): add Z-Image Base (undistilled) model variant support#8799
lstein merged 10 commits intoinvoke-ai:mainfrom
Pfannkuchensack:feature/z-image-undistilled-support

Pfannkuchensack commented Jan 28, 2026 •

edited

Loading

Uh oh!

lstein commented Jan 31, 2026 •

edited

Loading

Uh oh!

lstein commented Feb 1, 2026 •

edited by Pfannkuchensack

Loading

Uh oh!

Pfannkuchensack commented Feb 1, 2026

Uh oh!

lstein commented Feb 1, 2026

Uh oh!

skunkworxdark commented Feb 1, 2026

Uh oh!

Pfannkuchensack commented Feb 13, 2026 •

edited

Loading

Uh oh!

lstein left a comment

Uh oh!

Pfannkuchensack commented Feb 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Pfannkuchensack commented Jan 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Related Issues / Discussions

QA Instructions

Merge Plan

Checklist

Uh oh!

lstein commented Jan 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issues involving Z-Image Turbo support

Problem 1

Problem 2

Uh oh!

lstein commented Feb 1, 2026 • edited by Pfannkuchensack Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Pfannkuchensack commented Feb 1, 2026

Uh oh!

lstein commented Feb 1, 2026

Uh oh!

skunkworxdark commented Feb 1, 2026

Uh oh!

Pfannkuchensack commented Feb 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lstein left a comment

Choose a reason for hiding this comment

Uh oh!

Pfannkuchensack commented Feb 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Pfannkuchensack commented Jan 28, 2026 •

edited

Loading

lstein commented Jan 31, 2026 •

edited

Loading

lstein commented Feb 1, 2026 •

edited by Pfannkuchensack

Loading

Pfannkuchensack commented Feb 13, 2026 •

edited

Loading