Krea 2 Raw

'The undistilled base checkpoint of Krea 2, a 12.8B single-stream MMDiT conditioned on stacked Qwen3-VL-4B hidden states and working in the Qwen-Image VAE latent space. Krea ships it as the checkpoint to fine-tune, not to sample from.'

weights: krea/Krea-2-Raw ↗
org: Krea
modality: image
tasks: text-to-image
license: Krea 2 Community License
released: 2026-06-22
total params: 17.38B
model.arch: krea2

Components

role	model	params	size	dtype	trained
Transformer	Krea 2 SingleStreamDiT (Raw) krea/Krea-2-Raw/raw.safetensors SingleStreamDiT (ai-toolkit krea2/src/mmdit.py) The original-layout single file that ai-toolkit loads. The 321.6M fp32 params are the norm scales and the per-block modulation offsets. The repo also ships a Diffusers copy in transformer/ (Krea2Transformer2DModel, same parameter count) that ai-toolkit does not use.	12.82B 12,820,073,036	26.28 GB	bf16+fp32	yes
Text encoder	Qwen3-VL-4B-Instruct krea/Krea-2-Raw/text_encoder transformers.Qwen3VLModel Includes the vision tower. ai-toolkit loads Qwen/Qwen3-VL-4B-Instruct instead; every tensor is identical to this copy (compared tensor by tensor). The vision tower is dropped after loading for text-to-image training.	4.44B 4,437,815,808	8.88 GB	bf16	no
Tokenizer	Qwen2 tokenizer krea/Krea-2-Raw/tokenizer transformers.Qwen2Tokenizer	—	—	—	—
VAE	Qwen-Image VAE krea/Krea-2-Raw/vae diffusers.AutoencoderKLQwenImage The Qwen-Image VAE upcast to fp32 (every tensor equals Qwen/Qwen-Image vae/ once cast back to bf16). ai-toolkit loads the bf16 original from Qwen/Qwen-Image.	126.9M 126,892,531	508 MB	fp32	no
total		17.38B	35.67 GB

Latent space

spatial

8×

channels

patch

2×2

autoencoder: Qwen-Image VAE
pixels per token: 16×16
notes: The Qwen-Image VAE is a video VAE (Wan 2.1 lineage). Images are encoded as a single frame and latents are normalized per channel with the latents_mean / latents_std from its config.

input	latent (c×t×h×w)	tokens
1024×1024	16×1×128×128	4,096	ai-toolkit sample default
1280×1280	16×1×160×160	6,400	top of the time-shift range
2048×2048	16×1×256×256	16,384	Turbo README example

Architecture

Type: Single-stream MMDiT: text and image tokens run through the same blocks
Blocks: 28
Hidden size: 6144 (48 query heads × 128, 12 KV heads)
FFN size: 16384 (SwiGLU)
Text conditioning: Hidden states from 12 Qwen3-VL layers (2, 5, … 35), 2560-d each, fused by a 4-block text transformer and a learned layer mix, then projected to 6144
Timestep modulation: One shared projection for all blocks, plus a learned offset per block
Norm / position: QK RMSNorm, sigmoid-gated attention output, 3-axis RoPE (32/48/48 dims, θ 1000)
Objective: Rectified flow (target noise − clean), exponential time shift μ 0.5 → 1.15 over 256–6400 image tokens
Distillation: None (model_index is_distilled: false)
Recommended sampling: 52 steps, CFG 3.5 (Krea README)

In AI Toolkit

model.arch: krea2
UI label: Krea 2 (raw) (image)
model.name_or_path: krea/Krea-2-Raw
source: extensions_built_in/diffusion_models/krea2/krea2.py
extensions_built_in/diffusion_models/krea2/src/mmdit.py
extensions_built_in/diffusion_models/krea2/src/pipeline.py
extensions_built_in/diffusion_models/krea2/src/text_encoder.py
toolkit/models/v2/vae/qwen_image.py
toolkit/models/v2/text_encoders/qwen3_vl.py
extensions_built_in/diffusion_models/ui.tsx
extra UI sections: model.low_vram, model.layer_offloading

UI defaults

quantize / quantize_te: true / true (qfloat8)
low_vram: true
timestep_type: linear
network.conv: disabled (linear LoRA only)

Specifics

Gated repo: Accept the Krea 2 Community License on the Hub and set a Hugging Face token before training.
Transformer file: Downloads raw.safetensors from name_or_path (the file name comes from the repo name’s last segment). Override with model_kwargs.checkpoint_filename, or point name_or_path at a local .safetensors file or folder.
Text encoder source: Loads Qwen/Qwen3-VL-4B-Instruct (model_kwargs.text_encoder_path), not text_encoder/ from the Krea repo. The vision tower is dropped. Prompts are wrapped in Krea’s fixed system template, and the 34 template tokens are sliced off the output.
VAE source: Loads vae/ from Qwen/Qwen-Image (model_kwargs.vae_path), not the fp32 copy in the Krea repo.
Resolution: Buckets snap to multiples of 16 (8× VAE × 2×2 patch).
Timesteps: Training uses flowmatch with Krea’s resolution-dependent shift (use_dynamic_shifting, μ 0.5 at 256 px to 1.15 at 1280 px).
Prompt length: Capped at 512 tokens (model_kwargs.max_text_length).
Quantization: first, tmlp*, tproj*, txtmlp*, txtfusion.projector and last* stay in full precision.
LoRA target: SingleStreamDiT
Saving: LoRA keys are saved with the diffusion_model. prefix (ComfyUI layout). Full fine-tunes save a single safetensors of the SingleStreamDiT state dict.
Sampling: Built-in Euler sampler with the same time shift. CFG is zero-normalized: the sampler uses guidance_scale − 1, so guidance_scale 1 turns CFG off. low_vram tiles the VAE decode.
Metadata base version: krea2

Example config

not verified

job: extensionconfig:  name: "my_krea2_lora_v1"  process:    - type: "sd_trainer"      training_folder: "output"      device: cuda:0      network:        type: "lora"        linear: 32        linear_alpha: 32      save:        dtype: bf16        save_every: 250        max_step_saves_to_keep: 4      datasets:        - folder_path: "/path/to/images"          caption_ext: "txt"          caption_dropout_rate: 0.05          resolution: [512, 768, 1024]      train:        batch_size: 1        steps: 2000        gradient_checkpointing: true        noise_scheduler: "flowmatch"        timestep_type: "linear"        optimizer: "adamw8bit"        lr: 1e-4        dtype: bf16        cache_text_embeddings: true      model:        name_or_path: "krea/Krea-2-Raw"        arch: "krea2"        quantize: true        quantize_te: true        low_vram: true      sample:        sampler: "flowmatch"        sample_every: 250        width: 1024        height: 1024        guidance_scale: 4        sample_steps: 25        prompts:          - "a fox walking in the snow"