Tag: llama.cpp

All the articles with the tag "llama.cpp".

• My vibe-selling rig is a 13-inch tablet

  25 Jul, 2026

  Benchmarks from a 128 GB Strix Halo tablet: four models from a dense 27B to a 122B MoE on both AMD backends, a Poolside Laguna three days after its llama.cpp merge, and why parameter count is the wrong number.