Deployed footprint — runs on virtually any Apple device
End-to-end 1-bit weights across embeddings, attention projections, MLP projections, and LM head
MLX-native format (1-bit g128) with inline dequantization kernels — no FP16 materialization
Cross-platform companion: also available as GGUF Q1_0_g128 for llama.cpp

Resources

Google Colab — try Bonsai in your browser, no setup required
Whitepaper — for more details on Bonsai, check out our whitepaper
Demo repo — comprehensive examples for serving, benchmarking, and integrating Bonsai
Discord — join the community for support, discussion, and updates
1-bit kernels: MLX fork (Apple Silicon) · mlx-swift fork (iOS/macOS) · llama.cpp fork (CUDA + Metal)
Locally AI — we have partnered with Locally AI for iPhone support

Model Overview

Item	Specification
Parameters	1.7B (~1.4B non-embedding)
Architecture	Qwen3-1.7B dense: GQA (16 query / 8 KV heads), SwiGLU MLP, RoPE, RMSNorm
Layers	28 Transformer decoder blocks
Context length	32,768 tokens
Vocab size	151,936
Weight format	MLX 1-bit g128
Deployed size	0.27 GB (12.8x smaller than FP16)
1-bit coverage	Embeddings, attention projections, MLP projections, LM head
License	Apache 2.0

Quantization Format: 1-bit g128

Each weight is a single bit: 0 maps to −scale, 1 maps to +scale. Every group of 128 weights shares one FP16 scale factor.

MLX's quantization formats generally store both a scale and a bias per group: w = mlx_scale * bit + mlx_bias. To pack our scale-only 1-bit weights into this format:

mlx_scale = 2 * original_scale
mlx_bias  = −original_scale

This reconstructs −scale when bit=0 and +scale when bit=1. Because MLX stores two FP16 values per group (scale + bias) instead of one, the effective bits per weight is slightly higher than the GGUF format:

MLX 1-bit g128: 1.25 bpw (1 sign bit + two 16-bit values amortized over 128 weights)
GGUF Q1_0_g128: 1.125 bpw (1 sign bit + one 16-bit scale amortized over 128 weights)

Memory Requirement

Parameter memory only (weights and scales loaded into memory):

Format	Size	Reduction	Ratio
FP16	3.44 GB	—	1.0x
MLX 1-bit g128	0.27 GB	92.2%	12.8x
GGUF Q1_0_g128	0.24 GB	93.0%	14.2x

The model directory on disk is ~0.28 GB (~16 MB larger) because it also includes tokenizer, config, and other metadata files alongside the weights.

Best Practices

Generation Parameters

Parameter	Default	Suggested range
Temperature	0.5	0.5 -- 0.7
Top-k	20	20 -- 40
Top-p	0.9	0.85 -- 0.95
Repetition penalty	1.0
Presence penalty	0.0

System Prompt

You can use a simple system prompt such as:

You are a helpful assistant

Quickstart

MLX (Python)

Requires PrismML fork of MLX with 1-bit kernel support (upstream PR pending):

bash
1pip install mlx-lm 2pip install mlx @ git+https://github.com/PrismML-Eng/mlx.git@prism

python

from mlx_lm import load, generate

model, tokenizer = load("prism-ml/Bonsai-1.7B-mlx-1bit")

response = generate(
    model,
    tokenizer,
    prompt="Explain quantum computing in simple terms.",
    max_tokens=256,
)
print(response)

MLX Swift (iOS / macOS)

1-bit Bonsai 1.7B runs natively on iPhone and iPad via MLX Swift. Requires our mlx-swift fork with 1-bit kernels (upstream PR pending).

Throughput (MLX / Apple Silicon)

Platform	Backend	TG128 (tok/s)	FP16 TG (tok/s)	TG vs FP16	PP512 (tok/s)	FP16 PP512 (tok/s)
M4 Pro 48 GB	MLX (Python)	288	62	4.6x	1,759	1,585
M4 Pro 48 GB	llama.cpp Metal	250	65	3.8x	2,305	2,291
iPhone 17 Pro Max	MLX Swift	130	—	—	1,523	—

Citation

If you use 1-bit Bonsai 1.7B, please cite:

bibtex

@techreport{bonsai,
    title   = {Bonsai: End-to-End 1-bit Language Model Deployment
               Across Apple, GPU, and Mobile Runtimes},
    author  = {Prism ML},
    year    = {2026},
    month   = {March},
    url     = {https://prismml.com}
}

Contact

For questions, feedback, or collaboration inquiries: contact@prismml.com