Supported Models

Discover the state-of-the-art MLX-optimized models powering MindCraft Studio. To use these models in MindCraft Studio, copy the repository ID and open MindCraft Studio > Settings > Base Model > Import Models > Import from Remote , paste the repository ID and start import.

Z-Image Models

QuantizationLinks

Z-Image Turbo (4bit)

Fast 8-step generation with excellent quality.

4-bit

Z-Image Turbo (8bit)

Higher precision for improved fidelity.

8-bit

Z-Image Turbo (Full Weight)

Original unquantized model for maximum precision.

Native

FLUX.2 Models

QuantizationLinks

FLUX.2 [klein] (4B) (4bit)

Fastest 4-step generation with image editing capability.

4-bit

FLUX.2 [klein] (9B) (4bit)

High quality 4-step generation with image editing capability.

4-bit

FLUX.2 [Klein] 9B-KV (4bit)

FLUX.2 Klein 9B model optimized with KV cache quantization for faster generation.

4-bit

Ideogram 4 Models

QuantizationLinks

Ideogram 4 (4bit)

MLX 4-bit quantized Ideogram 4 text-to-image model with exceptional typography and prompt understanding.

4-bit

Ideogram 4 (FP8)

Official 8-bit floating point Ideogram 4 model balancing quality and memory efficiency.

FP8

Krea 2 Models

QuantizationLinks

Krea 2 Turbo (4bit)

Fast 8-step generation model based on Krea 2 architecture.

4-bit

Krea 2 Turbo (8bit)

Higher precision 8-bit quantized Krea 2 Turbo model.

8-bit

Lens-Turbo Models

QuantizationLinks

Lens-Turbo (6bit)

MLX-optimized distilled text-to-image model for fast 4-step generation.

6-bit

Lens-Turbo (Full Weight)

Original unquantized Microsoft Lens-Turbo model for maximum precision.

Native

ERNIE-Image Models

QuantizationLinks

ERNIE-Image-Turbo (4bit)

Baidu's open 8B text-to-image DiT model, distilled for fast 8-step generation with strong instruction following and text rendering.

4-bit

ERNIE-Image-Turbo (Full Weight)

Original unquantized Baidu ERNIE-Image-Turbo model for maximum precision.

Native

Qwen-Image Models

QuantizationLinks

Qwen-Image-2512 (6bit)

Next-generation vision-language model with exceptional understanding.

6-bit

Qwen-Image-2512 (8bit)

8-bit quantization for enhanced precision and image quality.

8-bit

Qwen-Image-2512 (Full Weight)

Original unquantized vision-language model.

Native

Qwen-Image-Edit (Full Weight)

Specialized model for precise image editing and manipulation.

Native

FLUX.1 Models

QuantizationLinks

FLUX.1 Schnell

The fastest model in the FLUX.1 family, providing high-quality results in very few steps.

4-bit

FLUX.1 Dev

Base version of the FLUX.1 model, suited for research and non-commercial exploration.

4-bit

FLUX.1 Kontext

Enhanced context and instruction following model.

4-bit

Looking for More Models?

Visit the mflux-community repository mirror to discover more MLX quantized and community models.

Explore mflux-community

Model not listed?

MindCraft Studio works with MLX-compatible Z-Image Turbo, FLUX and SDXL models. If you've found a model that works well, let us know and we'll add it!