Bonsai-4B-GGUF-1bit

End-to-end 1-bit language model for llama.cpp (CUDA, Metal, CPU)

14.1x smaller than FP16 | 4.2x faster on RTX 4090 | runs on any device

Highlights

Deployed footprint fits on virtually any device with a GPU
End-to-end 1-bit weights across embeddings, attention projections, MLP projections, and LM head
GGUF Q1_0_g128 format for 1-bit packing of weights with shared scales for each group (group size 128).
Cross-platform: CUDA (RTX/datacenter), Metal (Mac), Swift (iPhone/iPad), Android
MLX companion: also available as MLX 1-bit g128 for native Apple Silicon inference

Frontier Efficiency

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: llama.cpp fork (CUDA + Metal) · MLX fork (Apple Silicon) · mlx-swift fork (iOS/macOS)
Locally AI — we have partnered with Locally AI for iPhone support

Model Overview

Item	Specification
Parameters	4.0B (~3.6B non-embedding)
Architecture	Qwen3-4B dense: GQA (32 query / 8 KV heads), SwiGLU MLP, RoPE, RMSNorm
Layers	36 Transformer decoder blocks
Context length	32,768 tokens
Vocab size	151,936
Weight format	GGUF Q1_0_g128
Deployed size	0.57 GB (14.2x smaller than FP16)
1-bit coverage	Embeddings, attention projections, MLP projections, LM head
License	Apache 2.0

Quantization Format: Q1_0_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.

Effective bits per weight: 1.125 (1 sign bit + 16-bit scale amortized over 128 weights).

Memory Requirement

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

Format	Size	Reduction	Ratio
FP16	8.04 GB	—	1.0x
GGUF Q1_0_g128	0.57 GB	93.0%	14.2x
MLX 1-bit g128	0.63 GB	92.2%	12.8x

The GGUF file on disk is 0.57 GB (~6.4 MB larger) because the format embeds the tokenizer, chat template, and model metadata 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

llama.cpp (CUDA)

# Clone the PrismML fork of llama.cpp (includes Q1_0_g128 kernels)
git clone https://github.com/PrismML-Eng/llama.cpp
cd llama.cpp

# Build with CUDA support
cmake -B build -DGGML_CUDA=ON && cmake --build build -j

# Run inference
./build/bin/llama-cli \
    -m Bonsai-4B-Q1_0_g128.gguf \
    -p "Explain quantum computing in simple terms." \
    -n 256 \
    --temp 0.5 \
    --top-p 0.85 \
    --top-k 20 \
    -ngl 99

llama.cpp (Metal / macOS)

# Clone the PrismML fork of llama.cpp (includes Q1_0_g128 kernels)
git clone https://github.com/PrismML-Eng/llama.cpp
cd llama.cpp

# Build with Metal support (default on macOS)
cmake -B build && cmake --build build -j

# Run inference
./build/bin/llama-cli \
    -m Bonsai-4B-Q1_0_g128.gguf \
    -p "Explain quantum computing in simple terms." \
    -n 256 \
    --temp 0.5 \
    --top-p 0.85 \
    --top-k 20 \
    -ngl 99

llama.cpp Server

./build/bin/llama-server \
    -m Bonsai-4B-Q1_0_g128.gguf \
    --host 0.0.0.0 \
    --port 8080 \
    -ngl 99

Open the web UI at http://127.0.0.1:8080, or see our llama.cpp fork for more examples.

Cross-Platform Throughput

Platform	Backend	TG128 (tok/s)	FP16 TG (tok/s)	TG vs FP16	PP512 (tok/s)	FP16 PP512 (tok/s)
RTX 4090	llama.cpp CUDA	440	105	4.2x	18,135	16,310
M4 Pro 48 GB	llama.cpp Metal	136	29	4.7x	915	915

Citation

If you use 1-bit Bonsai 4B, please cite:

@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}
}