Instructions to use LocalMuseAI/mlx-flux2-klein-4b-aio-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use LocalMuseAI/mlx-flux2-klein-4b-aio-4bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download LocalMuseAI/mlx-flux2-klein-4b-aio-4bit --local-dir mlx-flux2-klein-4b-aio-4bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download source.json from LocalMuseAI/mlx-flux2-klein-4b-aio-4bit: direct link, hf CLI and curl.
- Browser
- Download file 3.14 kB
-
https://huggingface.co/LocalMuseAI/mlx-flux2-klein-4b-aio-4bit/resolve/main/source.json
- Command line
-
hf download hf://LocalMuseAI/mlx-flux2-klein-4b-aio-4bit/source.json
-
curl -L -o source.json https://huggingface.co/LocalMuseAI/mlx-flux2-klein-4b-aio-4bit/resolve/main/source.json
3.14 kB
| { | |
| "sourceURL": "https://civitai.com/models/2327389?modelVersionId=2618128", | |
| "modelId": 2327389, | |
| "modelVersionId": 2618128, | |
| "fileId": 2511488, | |
| "creator": { | |
| "username": "SeeSeeLP", | |
| "image": "https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1f8c6186-b6db-490a-a5d9-48b195e5f722/width=96/SeeSeeLP.jpeg" | |
| }, | |
| "modelName": "FLUX.2 [klein] AIO", | |
| "versionName": "Flux.2-4B-Distilled-AIO", | |
| "permissions": { | |
| "allowNoCredit": true, | |
| "allowCommercialUse": [ | |
| "Image", | |
| "RentCivit", | |
| "Rent", | |
| "Sell", | |
| "SellMerge" | |
| ], | |
| "allowDerivatives": true, | |
| "allowDifferentLicense": true | |
| }, | |
| "file": { | |
| "id": 2511488, | |
| "sizeKB": 15754216.890625, | |
| "name": "flux2KleinAIO_flux24BDistilledAIO_2511488.safetensors", | |
| "type": "Model", | |
| "pickleScanResult": "Success", | |
| "pickleScanMessage": "No Pickle imports", | |
| "virusScanResult": "Success", | |
| "virusScanMessage": null, | |
| "scannedAt": "2026-01-23T21:57:07.455", | |
| "metadata": { | |
| "format": "SafeTensor", | |
| "size": "full", | |
| "fp": "bf16" | |
| }, | |
| "hashes": { | |
| "AutoV1": "D6507209", | |
| "AutoV2": "F9F60D53BB", | |
| "SHA256": "F9F60D53BB21970E19D31D85EEFB294F9B0B5E36669EC81BA2BF52F102A7679E", | |
| "CRC32": "7CC572E7", | |
| "BLAKE3": "6ED985C247AC484B2A8D4157B90C2DCF608506D2C47F09F149893D6D41A31AAE", | |
| "AutoV3": "2C4856C66ED4", | |
| "SHA256_12": "F9F60D53BB21" | |
| }, | |
| "primary": false, | |
| "downloadUrl": "https://civitai.com/api/download/models/2618128?fileId=2511488" | |
| }, | |
| "baseModel": "Flux.2 Klein 4B", | |
| "baseRepository": { | |
| "repo": "black-forest-labs/FLUX.2-klein-4B", | |
| "revision": "e7b7dc27f91deacad38e78976d1f2b499d76a294" | |
| }, | |
| "supplementalVAE": { | |
| "repo": "black-forest-labs/FLUX.2-klein-4B", | |
| "revision": "e7b7dc27f91deacad38e78976d1f2b499d76a294", | |
| "file": "vae/diffusion_pytorch_model.safetensors", | |
| "sizeBytes": 168120878, | |
| "sha256": "ca70d2202afe6415bdbcb8793ba8cd99fd159cfe6192381504d6c4d3036e0f04", | |
| "usedTensors": [ | |
| "post_quant_conv.weight", | |
| "post_quant_conv.bias" | |
| ], | |
| "reason": "The converter looked up Diffusers' top-level post_quant_conv. This AIO nests the same learned projection as vae.decoder.post_quant_conv (BFL layout); its FP32 values round exactly to these official BF16 values, so the restored tensor is equivalent. An identity projection produces mosaic." | |
| }, | |
| "licenseStatement": "License: Apache 2.0 \u2013 Free for personal AND commercial use!", | |
| "authorDescriptionSummary": "All-in-One repack of the official Black Forest Labs Klein 4B step-distilled model, containing the Qwen3 text encoder, transformer and VAE. The author recommends 4\u20136 steps, CFG 1, Euler, Simple or normal scheduling, and native 1024x1024. Supports natural-language image generation; this app profile provides Text to Image and Image to Image reference editing.", | |
| "vaeEncoder": { | |
| "file": "vae_encoder.safetensors", | |
| "source": "AIO vae.encoder.* in FP32, including the nested learned vae.encoder.quant_conv", | |
| "use": "Image to Image reference editing (Diffusers Flux2KleinPipeline reference tokens, RoPE time 10)" | |
| } | |
| } | |