Related paper

BitNet: Scaling 1-bit Transformers for Large Language Models (2023): https://arxiv.org/abs/2310.11453

This came out too close to April 1 for me to know if it was real or not :).

I do like the idea of local/on device llms so it's interesting and will keep an eye on it.