SceneWorks/z-image-mlx

1---2license: apache-2.03language:4- en5pipeline_tag: text-to-image6library_name: mlx7tags:8- mlx9- apple-silicon10- diffusion11- z-image12- text-to-image13- quantized14base_model: Tongyi-MAI/Z-Image15---16 17# Z-Image — MLX quant-matrix (SceneWorks re-host)18 19Pre-built **MLX** (Apple Silicon) quantization tiers of [`Tongyi-MAI/Z-Image`](https://huggingface.co/Tongyi-MAI/Z-Image), hosted by [SceneWorks](https://github.com/SceneWorks) for direct, ready-to-run loading in the SceneWorks desktop app (no install-time conversion, no gated download).20 21## Tiers22 23Each subdirectory is a complete, self-contained snapshot (transformer + Qwen3 text encoder + VAE + tokenizer + scheduler) that the SceneWorks `z_image` engine loads directly:24 25| Tier | Subdir | Precision | Use |26|------|--------|-----------|-----|27| Q4 (default) | `q4/` | 4-bit group-affine (group 64) weights; dense norms | smallest footprint (undistilled base, real CFG) |28| Q8 | `q8/` | 8-bit group-affine weights | higher fidelity |29| bf16 | `bf16/` | dense bf16 | maximum fidelity |30 31The transformer, text encoder, and VAE attention are quantized in the Q4/Q8 tiers; the bf16 tier is the full dense model. The packed weights auto-detect their quantization on load (no manifest needed).32 33## License34 35Apache-2.0, inherited from the upstream `Tongyi-MAI/Z-Image`. This is an unmodified-weights re-host (re-quantized for MLX). All credit to the Tongyi-MAI team.36