Can Ternary Bonsai 27B run on MacBook Air M3 24GB?

YES — Runs great

Q2_0 · 4K context · formula estimate

What this means

Ternary Bonsai 27B fits on the MacBook Air M3 · 24GB at Q2_0. We estimate it uses 8.6GB of the conservative 21GB working budget.

Estimated decode speed is 5.8–8.1 tokens/s. A roughly 300-word answer may take around 57 seconds.

View technical details ↓Setup guide →

5.8–8.1tokens/s

estimated · reading pace

8.6GB

needed at Q2_0

Q2_0

quant selected

12.4GB

working headroom

needs 8.6 GBworking budget 21 GB

Apple M324 GB unified memoryMetalFanless

See how fast it feels

Using the midpoint of our 5.8–8.1 tokens/s estimate for this demo.

Live demo · 7 tokens/s

▌

Compare other speeds →

Where the memory goes

Component	Detail	GB
Model weights	Q2_0 GGUF (7.2 GB) + mmap overhead	7.6
KV cache	4K context window	0.3
Runtime	macOS inference app + compute buffers	0.8
Total needed	Q2_0, 4K context	8.6
Working budget	24 GB unified memory − conservative macOS reserve	21
Headroom	remaining inside the working budget	12.4

Pick your quant

Quant	Download	Memory	Estimated speed	Verdict
Q2_0 ★	7.2 GB	8.6 GB	~5.8–8.1 tok/s	✓ Runs great
Q2_g64	7.6 GB	9 GB	~5.5–7.6 tok/s	✓ Runs great

Get it running on this Mac

Choose the beginner-friendly build instead.

This Mac has enough memory, but the tracked Q2_0 ternary file is not supported by the current point-and-click apps. For a first local chat, use Bonsai 27B 1-bit in Locally AI. Do not download PQ2_0; the publisher marks it unsupported.

Only use the publisher tool if you are comfortable with Terminal.

The advanced path is complete below, but it is not required for ordinary local AI.

Advanced publisher setup (Terminal)

Install Apple's command-line tools if macOS asks, then run:

git clone https://github.com/PrismML-Eng/Bonsai-demo.git
cd Bonsai-demo
BONSAI_OPENWEBUI=0 BONSAI_SKIP_MLX=1 BONSAI_FAMILY=ternary BONSAI_MODEL=27B ./setup.sh
BONSAI_FAMILY=ternary BONSAI_MODEL=27B BONSAI_CTX=4096 ./scripts/start_llama_server.sh

When the server says it is ready, open http://localhost:8080. Start with 4K context. This route uses the publisher's Prism runtime, not Jan or LM Studio.

Other models on this Mac

Qwen 3.6 35B-A3B · IQ4_XS Qwen 3.6 27B · Q4_K_M Bonsai 27B (1-bit) · Q1_0 Gemma 4 26B-A4B · Q4_0 GPT-OSS 20B · MXFP4 Qwen3 14B · Q4_K_M

Ternary Bonsai 27B on other MacBook Air M3 configurations

Check 16GB

FAQ

Can the MacBook Air M3 · 24GB run Ternary Bonsai 27B?

Yes at Q2_0. We estimate about 8.6GB of working memory and 5.8–8.1 tokens/s at 4K context.

Which Ternary Bonsai 27B quant should I use on this Mac?

Q2_0. It is a 7.2GB download and leaves about 12.4GB inside our conservative working budget.

Which app should I use for Ternary Bonsai 27B on this Mac?

Current beginner Mac apps do not load this exact file. Use Bonsai 27B 1-bit in Locally AI instead, or open the advanced publisher instructions on this page.

Are these speeds measured on a MacBook Air M3?

No. The range is a formula estimate based on memory bandwidth, model size, active parameters, cooling, and a 4K context. The app, backend, thermals, and prompt can change real performance.