Text-to-Image
MLX
Safetensors

FLUX 3 Image Klein MLX Q4

Unofficial affine 4-bit / group-64 derivative. Complete package includes Q4 DiT and Qwen3-VL text weights, FP32 VAE, tokenizer and original license/notices. Native Swift/MLX inference: zimageapp/zimage, Flux3CLI --model /path/to/package --prompt "a red apple" --output out.png.

Default: 1024 square, 16 distilled flow-map steps, no CFG. --fp16-dense selects attention projections; --cache-di-t enables approximate fixed-interval prediction reuse. Both are opt-in.

The original FLUX Kommunity License applies directly to recipients. See LICENSE.md and NOTICE.md; conversion.json records source revision, hashes and quantization details. This derivative is not endorsed by Black Forest Labs. Model implementation and performance reports are in the source repository.

Downloads last month
52
Safetensors
Model size
6B params
Tensor type
U32
·
BF16
·
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for zimageapp/FLUX-3-Klein-MLX-Q4

Finetuned
(10)
this model