Ternary Bonsai 2 27B runs at 47 tok/s (46.8 as recorded) on an M5 Max, a published figure reported by github.com (llama.cpp, PQ2_0, tg128, depth 0, batch 1). It needs about 9 GB of unified memory. It is Prism ML's 27.36B dense model, released under the Apache 2.0 licence.
Want to run Ternary Bonsai 2 27B faster — or step up to bigger models? Find your Mac →This is a published figure: a real run, reported by its source and linked below. The index records it as given and never adjusts it.
| Metric | Value |
|---|---|
| Tokens per second | sourced47 tok/s |
| Provenance | published figure · github.com |
| Quantization | PQ2_0 |
| Engine | llama.cpp |
| Context | tg128, depth 0, batch 1 |
| Figure dated | 2026-09-16 |
| Minimum RAM | 9 GB |
| Parameters | 27.36B |
| Chip bandwidth | 614 GB/s |
Each row as the index holds it, with its source. “Mac RAM” is the memory of the Mac a run used, not the model’s requirement (9 GB). The headline figure follows the reference rule: the lower-median measured plain-decode run at 5 bits or fewer (a run whose quant was not stated counts only when no stated-quant run qualifies), else the estimate; a run marked “not used to anchor estimates” is listed here but never becomes the headline.
| Speed | Quant | Engine | Mac RAM | Context | Source | Note |
|---|---|---|---|---|---|---|
| sourced46.8 tok/s | PQ2_0 | llama.cpp | — | tg128, depth 0, batch 1 | github.com | Vendor whitepaper Table 6: PrismML llama.cpp fork (Metal), release build, mean of 3 after warm-up; 7.21 GB. Replaces the model card's 47.0, which was labelled pre-rotation. |
Ollama — install it (brew install ollama or the app from ollama.com), then pull the model by its tag, checked on ollama.com on 2026-10-17:
MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:
Downloads: Ollama page → · model card →
The 47 tok/s figure above came from llama.cpp at PQ2_0; other runtimes and quants run at different speeds.
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
| Chip | Speed | Quant | Engine | Bandwidth |
|---|---|---|---|---|
| M5 Pro | sourced28 tok/s | PQ2_0 | llama.cpp | 307 GB/s |
Prism ML's ternary (1.58-bit) rebuild of Qwen3.8-27B (Sep 2026, Apache 2.0): ~7 GB, image input, 262K context. It keeps 98% of the base model's average on Prism's 14-benchmark suite but drops on agentic coding (SWE-bench Verified 60.8 vs 80.6 on Prism's scaffold). Ternary decoding is kernel-bound, so the index quotes only vendor figures: 46.8 tok/s on an M5 Max and 28.1 on an M5 Pro (PrismML llama.cpp fork, PQ2_0); Ollama's bonsai-2:27b ships the 8.3 GB Q2_0 pack.
Apache 2.0 licence · 27.36B · 262K context · LLMCheck capability 37/50 · see its leaderboard rank →
Ternary Bonsai 2 27B needs about 9 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep the LLMCheck index free.
See how Ternary Bonsai 2 27B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard