shieldstackllc/GLM-4.7-Flash-PRISM-mlx-8bit

1---2language:3- en4- zh5license: other6license_name: glm-4-license7pipeline_tag: text-generation8tags:9- mlx10- glm411- moe12- prism13- abliterated14- 8bit15- quantized16- apple-silicon17library_name: mlx18base_model: Ex0bit/GLM-4.7-Flash-PRISM19---20 21<p align="center">22  <a href="https://vmlx.net">23    <img src="vmlx-logo.png" alt="vMLX" width="120">24  </a>25</p>26 27# GLM-4.7-Flash-PRISM — MLX 8-bit28 29MLX 8-bit quantized version of [Ex0bit/GLM-4.7-Flash-PRISM](https://huggingface.co/Ex0bit/GLM-4.7-Flash-PRISM) for efficient local inference on Apple Silicon.30 31- **Quantization**: 8-bit (8.5 bits per weight, group size 64, affine mode)32- **Architecture**: GLM-4 MoE Lite — 47 layers, 64 routed experts, 4 active per token33- **Context**: 202K tokens34- **Size**: ~30 GB35 36## Usage37 38```python39from mlx_lm import load, generate40 41model, tokenizer = load("shieldstackllc/GLM-4.7-Flash-PRISM-mlx-8bit")42response = generate(model, tokenizer, prompt="Hello!", verbose=True)43```44 45Or with [vMLX](https://vmlx.net) for native macOS inference.46 47## About48 49This model is an abliterated (uncensored) variant of GLM-4.7-Flash, a Mixture-of-Experts language model by Zhipu AI / THUDM. The abliteration was done by [Ex0bit](https://huggingface.co/Ex0bit) as part of the PRISM series. MLX quantization by [vMLX](https://vmlx.net).50 51## Also Available52 53- [GLM-4.7-Flash-PRISM MLX 4-bit](https://huggingface.co/shieldstackllc/GLM-4.7-Flash-PRISM-mlx-4bit) (~16 GB)54 55## Made for vMLX56 57This model was converted and optimized for [vMLX](https://vmlx.net) — a free, open source macOS native MLX inference engine for Apple Silicon. Download vMLX to run this model locally with zero configuration.58 59## Credits60 61- **Base model**: [THUDM/GLM-4](https://github.com/THUDM/GLM-4) by Zhipu AI62- **Abliteration**: [Ex0bit/GLM-4.7-Flash-PRISM](https://huggingface.co/Ex0bit/GLM-4.7-Flash-PRISM)63- **MLX conversion**: [vMLX](https://vmlx.net) — Run AI locally on Mac. No compromises.64 65## Contact66 67For questions, issues, or collaboration: **admin@vmlx.net**68