Modelpublic

TyKaoz/Apertus-v1.5-8B-text-6bit

sourceHugging Faceapache-2.0updated 8d agoView on Hugging Face

0likes217downloads

Model Card

Apertus v1.5 8B text — 6-bit MLX

6-bit MLX quantization of `swiss-ai/Apertus-v1.5-8B`, for Apple Silicon (~6.1 GB). Text branch of the omni checkpoint (the image and audio towers and codebooks are not included), extracted from the original weights and quantized with mlx-lm. xIELU activation, 262144-token context, the source tokenizer and chat template. Apertus is the fully open model of the Swiss AI Initiative (EPFL, ETH Zurich, CSCS).

Usage

bash

pip install -U mlx-lm

bash

mlx_lm.generate \
  --model TyKaoz/Apertus-v1.5-8B-text-6bit \
  --prompt "Explique la quantization en une phrase." \
  --max-tokens 200

Base	Tool	Precision	Size
`swiss-ai/Apertus-v1.5-8B`	`mlx-lm`	6-bit · group 64	~6.1 GB

By [TyKaoz](https://www.tykaoz.bzh) — privacy-first native macOS LLM chat client. Apache 2.0, inherited from the base model, with the Apertus acceptable use policy.