Post
LR
lr.duangle.com
did:plc:e27fp6a6vcarkkbwbl5niuex
okay, the whole thing got fairly easy to set up now
https://github.com/PrismML-Eng/llama.cpp/releases/tag/prism-b10777-e3de715
+
https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf/blob/main/Ternary-Bonsai-2-27B-PTQ1_0.gguf
35 t/s on my RTX 3060 12GB
see also https://huggingface.co/ukisai/Swift-Bonsai-2-GGUF/blob/main/Swift-Bonsai-2-PTQ1_0.gguf which fixes overthinking
2026-10-15T09:43:04.553Z