This is a decensored version of prism-ml/Ternary-Bonsai-27B-unpacked, made using Heretic v1.4.0 with the Arbitrary-Rank Ablation (ARA) method using a LoRA adapter and row-norm preservation
Abliteration parameters
| Parameter | Value |
|---|
| start_layer_index | 10 |
| end_layer_index | 51 |
| preserve_good_behavior_weight | 0.6491 |
| steer_bad_behavior_weight | 0.0008 |
| overcorrect_relative_weight | 0.9874 |
| neighbor_count | 14 |
Performance
Note: Performance testing, including the measurement of refusal rates, was conducted using Japanese datasets (
harmless_alpaca_ja,
harmful_behaviors_ja).
GGUF Version
GGUF quantizations available
here⚠️ Important Notice
This model has undergone substantial reduction of its safety alignment. As a result, it is more likely than standard models to generate harmful, inaccurate, biased, offensive, or otherwise inappropriate content.
Intended Use
For research and experimentation only, including safety research, alignment studies, and red-teaming. Please avoid deploying it in public or end-user-facing services.
User Responsibility
All outputs should be treated as untrusted and independently verified before use. Users are solely responsible for:
- Evaluating the accuracy and suitability of generated content
- Implementing appropriate safeguards and human oversight
- Complying with applicable laws, regulations, licenses, and ethical standards
Use of this model is entirely at your own risk.
Disclaimer
OS-Software provides this model without warranties of any kind and assumes no liability for any direct or indirect damages, losses, misuse, or legal consequences arising from its use.
Acknowledgements
Thanks to the base model developers,
p-e-w for Heretic, and the wider open-source community.
This is a derivative work released under the base model’s applicable license. All rights to the base model remain with their respective owners.
Ternary Bonsai 27B — Unpacked FP16 Safetensors
FP16 safetensors (HuggingFace format) of the Ternary Bonsai 27B model. This repo exists for users who want to run Ternary Bonsai with stock HuggingFace tooling or frameworks that don't yet support the packed ternary format. The 2-bit hybrid-attention kernels are currently in our forks of
MLX,
mlx-swift, and
llama.cpp — once they land upstream, this unpacked version will no longer be needed.
We strongly recommend using the natively packed models instead. The packed format is where all the benefits of Bonsai come from — a 7.2 GB deployed footprint (down from 54 GB), 95% of FP16 intelligence retained, and interactive decoding on everyday laptops (26 tok/s on an M5 Pro). This unpacked FP16 version is full-size and does not provide any of those advantages.
For the optimized ternary release models (recommended):
- Ternary-Bonsai-27B-mlx-2bit — Ternary MLX for Apple Silicon
- Ternary GGUF (Q2_0_g128) for llama.cpp (CUDA, Metal, CPU)
For the phone-class variant:
- Bonsai-27B-mlx-1bit — 1-bit Bonsai 27B (~3.9 GB), fits an iPhone 17 Pro Max
