| # | Tweet | Community | Topic | Views ▼ | Ratio | Engagement | Posted |
|---|---|---|---|---|---|---|---|
| 1 | [image] 다음 목표는 현 시점에서 소비자 레벨에서 가장 무난한 모델 중 하나인 gemma-4-26b를 128k 이상 컨텍스트 윈도우를 뽑으면서 m4 맥미니 16gb에서 30tps 이상 나올 수 있도록 양자화 튜닝하는 것 입니다. | OpenClaw | — | 32.6K | 0.7x | 128 | Apr 8 |
| 2 | [image] Gemma-4 가 나왔습니다. | OpenClaw | — | 18.9K | 0.4x | 78 | Apr 2 |
| 3 | [image] The smallest full-scale GLM-5.2 build on MLX has come out.
You can download it now, and it’s also so good.
It uses 2.32 bits per weight (219 GB) and fits on a 256 GB Mac with room left for context.
Models this far down are usually barely usable.
An anchor-guarded clip | Local AI | — | 15.3K | 0.2x | 67 | Jul 12 |
| 4 | [image] This is the third version of KIMI-K3-alis-mlx. Compared to the previous V2, the weight size has been reduced by 211GB (from 948.5GB to 737.1GB), and the quality has been improved to be closer to the original through the DWQ technique.
It still cannot run on a single M3 ultra Mac | Local AI | — | 13.4K | 0.2x | 75 | Aug 4 |
| 5 | [image] Introducing MOTIF-3-alis-mlx-4.5bpw.
This build maximizes speed on MLX platforms with 256GB or more without significantly compromising quality.
Based on M3 Ultra, it achieves a prefill speed of 790 tok/s and a decode speed (with MTP) of 50 tok/s.
*MOTIF-3 is currently the | x/LocalLLaMA | — | 9.0K | 0.1x | 52 | Aug 14 |
