Ternary Bonsai 2 27B · llama.cpp (PrismML fork) · MacBook M5 Pro
English | 简体中文 | 日本語 · ← All recipes
Source level: 📣 Claimed · Date: 2026-09-16 · Numbers come from the source. This repo did not measure them again.
Hardware
- MacBook M5 Pro (Mac)
- Reference cost: Not published. See the source.
Model
- Ternary Bonsai 2 27B ·
PQ2_0 - Download: huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf
Engine
- llama.cpp (PrismML fork) — install: https://github.com/ggml-org/llama.cpp
- Note: PrismML llama.cpp fork (Metal)
Launch command
llama-cli ... -ngl 99 -fa on -c 32768 (requires the PrismML llama.cpp fork; stock llama.cpp does not work)
Excerpt from the source (includes ... or placeholders). See the source for full arguments.
Key parameters
- Context: 262K (model card); example command -c 32768
- Concurrency: batch 1 (single stream)
- Tags:
ternary,dense,long-context
Expected speed
- 28.1 tok/s decode; 387 tok/s prefill (📣 Claimed)
Caveats
- Not a first-hand measurement (claimed or relayed). Your results can differ.
- Speed changes with drivers, engine version, context length, and prompt length.
Source & attribution
- https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf
- Credit to the original author. This page only summarizes and links. Data under CC BY 4.0.
Source file: recipes/hf-prism-ml-ternary-bonsai-2-27b-gguf-5.md