bonsai-27b-2bit
Run Ternary Bonsai 27B on a Mac
A 27 B model in 7.9 GB of weights — ternary 2-bit quantisation makes a 27 B fit where a 4-bit 12B sits. Decode is slower per token than MoE peers, but the quality-per-gigabyte is the point.
27 B ternary · 2-bitbiggest model that fits on a 16 GB Mac
256K context
One command
rapid-mlx serve bonsai-27b-2bit
Weights (7.9 GB) download on first run; you get an OpenAI-compatible
endpoint at http://localhost:8000/v1. No rapid-mlx yet? It's one line —
curl -fsSL https://rapidmlx.com/install.sh | bash — or take the
desktop app.
Point a client at it. OpenAI SDK: base_url="http://localhost:8000/v1",
model="bonsai-27b-2bit". Claude Code:
ANTHROPIC_BASE_URL=http://localhost:8000 (no /v1 suffix) —
see the Claude Code guide.
Measured on real hardware
Headline machine · Mac mini M4 Pro · 48 GB
All measured machines
| Machine | Decode | First token | Peak memory | Cold boot | Weights |
|---|---|---|---|---|---|
| Mac mini M4 Pro · 48 GB rapid-mlx 0.15.5 · 2026-10-03 raw data | 22.4 tok/s | 19.4 s | 13.5 GB | — | 7.9 GB |
Each row cites its own run; numbers are never copied across machines. Mac mini M4 Pro · 48 GB: rapid-mlx 0.15.5, 2026-10-03 — raw data + method — 4 concurrent streams: 35.8 tok/s aggregate.
Community runs, one page per Mac: Apple M2.
Will it fit your Mac?
Measured 13.5 GB Metal peak (7.7 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
For the conservative install-default placement see the hardware tiers table; to compare against every model your RAM can hold, use the live picker.
Variants & profiles
gemma-4-26b-4bit— Gemma 4 26B MoE — similar class, much faster decode, more memory
FAQ
How much memory does Ternary Bonsai 27B need on a Mac?
Measured 13.5 GB Metal peak (7.7 GB process RSS) on a 48 GB M4 Pro across the whole benchmark run — up to a ~16K-token prompt and four concurrent streams, including the engine's prefix cache, which defaults to 20% of RAM. A short chat needs less; long contexts grow the KV cache, so leave headroom.
How fast is Ternary Bonsai 27B on Apple Silicon?
We measured 22.4 tok/s on a Mac mini M4 Pro · 48 GB, rapid-mlx 0.15.5, 2026-10-03; peak memory 13.5 GB on that run. Every row on this page keeps its own chip, version and date — no number is copied across machines.
How do I run Ternary Bonsai 27B locally?
rapid-mlx serve bonsai-27b-2bit — the weights download on first run and you get an OpenAI-compatible endpoint at localhost:8000/v1 for Aider, Codex and any OpenAI client, plus the Anthropic API at localhost:8000 for Claude Code.