Files read on Hugging Face: 2026-10-24

Krea 2 Turbo on Mac: files, RAM and published times

Krea 2 Turbo is the 8-step distilled text-to-image checkpoint of Krea 2. Commercial use is allowed only below USD $1,000,000 in annual revenue under the Krea 2 Community License.

Krea · Updated 2026-10-24
Answer

Krea 2 Turbo runs on a Mac with 32 GB or more. A published run on M1 Max · 32 GB used the Turbo MLX mixed-4/8 build (source, 2026-07-08).

Smallest Mac with a published run
32 GBM1 Max · 32 GB · Turbo MLX mixed-4/8
Smallest complete download
16.87 GBTurbo Q4_K_M GGUF
Files read
2026-10-24Hugging Face

Krea 2 Community License Agreement v.1. Commercial use: restricted. Excluded territories: none listed. Commercial use needs total company-wide annual revenue below USD $1,000,000 on a trailing twelve-month basis, or an Enterprise License.

See the exact GGUF files and sizes

Builds here are Krea 2 Turbo, the 8-step distilled checkpoint. Krea 2 Raw is a separate undistilled checkpoint with its own 52-step settings, and Turbo results do not apply to it.

Table 1

What do you download?

Only builds with every required component and an exact probed file size are shown.

Turbo MLX mixed-4/8

mlx · mixed-4/8 · 19.22 GB complete · Recommended

Mac routes: ComfyUI-Krea2-MLX

Exact component files for Turbo MLX mixed-4/8
FileComponentRequiredDownloadSource
transformer_mixed_4_8.safetensorsavlp12/Krea-2-Turbo-Alis-MLX-mixed-4-8Krea 2 transformerditYes9.84 GBHugging FaceRead 2026-10-24
text_encoder/model.safetensorskrea/Krea-2-TurboQwen3-VL-4B-Instruct text encodertext_encoderYes8.88 GBHugging FaceRead 2026-10-24
vae/diffusion_pytorch_model.safetensorskrea/Krea-2-TurboQwen-Image VAEvaeYes0.51 GBHugging FaceRead 2026-10-24

Build sources: ComfyUI-Krea2-MLX README.

Turbo Q4_K_M GGUF

gguf · Q4_K_M · 16.87 GB complete

Mac routes: stable-diffusion.cpp

Exact component files for Turbo Q4_K_M GGUF
FileComponentRequiredDownloadSource
krea2_turbo-Q4_K_M.ggufvantagewithai/Krea-2-Turbo-GGUFKrea 2 transformerditYes7.49 GBHugging FaceRead 2026-10-24
text_encoder/model.safetensorskrea/Krea-2-TurboQwen3-VL-4B-Instruct text encodertext_encoderYes8.88 GBHugging FaceRead 2026-10-24
vae/diffusion_pytorch_model.safetensorskrea/Krea-2-TurboQwen-Image VAEvaeYes0.51 GBHugging FaceRead 2026-10-24

Build sources: Krea 2 Turbo GGUF card · stable-diffusion.cpp README.

Turbo Q8_0 GGUF

gguf · Q8_0 · 23.09 GB complete

Mac routes: stable-diffusion.cpp

Exact component files for Turbo Q8_0 GGUF
FileComponentRequiredDownloadSource
krea2_turbo-Q8_0.ggufvantagewithai/Krea-2-Turbo-GGUFKrea 2 transformerditYes13.71 GBHugging FaceRead 2026-10-24
text_encoder/model.safetensorskrea/Krea-2-TurboQwen3-VL-4B-Instruct text encodertext_encoderYes8.88 GBHugging FaceRead 2026-10-24
vae/diffusion_pytorch_model.safetensorskrea/Krea-2-TurboQwen-Image VAEvaeYes0.51 GBHugging FaceRead 2026-10-24

Build sources: Krea 2 Turbo GGUF card · stable-diffusion.cpp README.

Turbo original safetensors

safetensors · Original weights · 35.67 GB complete

Mac routes: Diffusers

Exact component files for Turbo original safetensors
FileComponentRequiredDownloadSource
transformer/diffusion_pytorch_model-00001-of-00003.safetensorskrea/Krea-2-TurboKrea 2 transformerditYes9.97 GBHugging FaceRead 2026-10-24
transformer/diffusion_pytorch_model-00002-of-00003.safetensorskrea/Krea-2-TurboKrea 2 transformerditYes9.83 GBHugging FaceRead 2026-10-24
transformer/diffusion_pytorch_model-00003-of-00003.safetensorskrea/Krea-2-TurboKrea 2 transformerditYes6.48 GBHugging FaceRead 2026-10-24
text_encoder/model.safetensorskrea/Krea-2-TurboQwen3-VL-4B-Instruct text encodertext_encoderYes8.88 GBHugging FaceRead 2026-10-24
vae/diffusion_pytorch_model.safetensorskrea/Krea-2-TurboQwen-Image VAEvaeYes0.51 GBHugging FaceRead 2026-10-24

Build sources: Krea 2 Turbo model card.

Table 2

Does it fit your Mac's RAM?

Published names the exact measured machine. est. is a memory calculation. unknown means the run verdict is not established, not that the model fails.

Run verdicts by Mac memory tier
Mac RAMComplete buildDownloadRun verdictBasis
8 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBDoes not fitest.
16 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBDoes not fitest.
24 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBunknownunknown
32 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBRuns · PublishedPublishedApple M1 Max · 32 GB RAM
48 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBRuns · Published on M1 Max · 32 GBPublishedApple M1 Max · 32 GB RAM
64 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBRuns · Published on M1 Max · 32 GBPublishedApple M1 Max · 32 GB RAM
96 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBRuns · Published on M1 Max · 32 GBPublishedApple M1 Max · 32 GB RAM
128 GBTurbo MLX mixed-4/8mixed-4/8 · ComfyUI-Krea2-MLX19.22 GBRuns · Published on M1 Max · 32 GBPublishedApple M1 Max · 32 GB RAM

How does it rank in blind votes?

No independent score yet.

Table 3

What published Mac times are available?

Third-party results keep their original machine and settings. ModelFit does not run these benchmarks.

Published Mac generation measurements
Machine and RAMSettingsAppResultSource
Apple M1 Max · 32 GB RAM1024x1024 · 8 steps · mixed-4/8 · ComfyUI-Krea2-MLX v0.4.0 with the Turbo mixed-4/8 build. Sampler, guidance and run date are not stated; the date is the README commit date.ComfyUI-Krea2-MLXSuccess~263sPeak memory unknownComfyUI-Krea2-MLX READMEPublished 2026-07-08 · Grade A

Which Mac software routes are listed?

  • ComfyUI-Krea2-MLX

    Community · v0.4.0

    Formats: mlx. Memory policy: component stages stay resident.

    Unofficial ComfyUI node pack that runs Krea 2 Turbo in-process on MLX. It keeps one model resident and does not plug into the native KSampler.

    Sources: ComfyUI-Krea2-MLX README

  • Diffusers

    Official

    Formats: safetensors, gguf. Memory policy: component stages stay resident.

    Runs on the PyTorch mps device on Mac and CUDA on NVIDIA. GGUF loads through from_single_file on the model class, not the pipeline. Resident size is the sum of the weights, an upper bound.

    Sources: Diffusers MPS guide · Diffusers GGUF guide

  • mflux

    Community · 0.23.0

    Formats: safetensors. Memory policy: component stages stay resident.

    MLX-native runner. It reads the original Hugging Face weights and quantizes them while loading (--quantize 3, 4, 5, 6 or 8). --low-ram lowers memory use at a performance cost.

    Install with uv

    uv tool install --upgrade mflux

    Sources: mflux README, install · mflux pyproject · mflux common README, quantization

  • stable-diffusion.cpp

    Official

    Formats: gguf, safetensors. Memory policy: component stages stay resident.

    C/C++ runner with Metal on Mac and CUDA on NVIDIA. It reads GGUF and safetensors, and its docs keep a page per model family. Resident size is the sum of the weights, an upper bound.

    Sources: stable-diffusion.cpp README, backends

What does the official configuration say?

"Official Turbo settings are 8 steps, cfg 0.0 and mu 1.15, with generation from 1k to 2k resolution. No numeric RAM or VRAM minimum is given."

Source: Krea 2 README · Read 2026-10-24.

What known issues should you check?

  • Known issue 1

    The node author describes one 12.9B model as staying resident at about 18 to 20 GB with encoder and VAE. That is an estimate, not a measured peak.

    Source: ComfyUI-Krea2-MLX README · Read 2026-10-24

  • Known issue 2

    The Mac time comes from the node author README, with sampler, guidance and run date not stated. The source date is the commit that published it.

    Source: ComfyUI-Krea2-MLX README · Read 2026-10-24

Frequently asked questions

How fast is Krea 2 Turbo on a Mac?

The ComfyUI-Krea2-MLX author reports about 263 s for a 1024x1024 image in 8 steps with the mixed-4/8 build on an M1 Max with 32 GB. Sampler, guidance and run date are not stated, so the date shown is the README commit.

Can I use Krea 2 commercially?

Only below a revenue limit. The Krea 2 Community License allows commercial use when total company-wide annual revenue is under USD $1,000,000 on a trailing twelve-month basis. Above that you need an Enterprise License.

What is the difference between Krea 2 Turbo and Raw?

Turbo is the 8-step distilled checkpoint, with official settings of 8 steps and cfg 0.0. Raw is a separate undistilled checkpoint with its own 52-step settings, and Turbo results do not apply to it.

Sources