LFM2.5-8B-A1B-uncensored-abliterated
AdvBench:
| refusals (n=100) | |
|---|---|
| original | 39/100 |
| this | 0/100 |
Method
- Dual-path abliteration targeting LFM2.5 hybrid architecture (18 conv + 6 attention layers)
- 50 modules modified across all 24 layers (conv + attention paths)
- Refusal direction: harmful - harmless
- Per-module weights: attn=4.0, conv=3.0, ffn=2.5, in_proj=2.0
- 51 prompts for precise direction estimation
License: LFM 1.0 License
Base Model Card (LiquidAI/LFM2.5-8B-A1B)
LFM2.5-8B-A1B
LFM2.5 is a new family of hybrid models designed for on-device deployment. It builds on the LFM2 architecture with extended pre-training and reinforcement learning.
- On-device personal assistant: Designed to power real-life applications, chaining tool calls, and following complex instructions on all devices.
- Compressed performance: Competitive with much larger dense and MoE models on instruction following and agentic tasks.
- Unmatched throughput: Fastest in its size class on both CPU and GPU inference, with day-one support for llama.cpp, MLX, vLLM, and SGLang.
Find more information about LFM2.5-8B-A1B in our blog post.
Model Details
- Model size: 8B params
- Tensor type: F32 / BF16
- Architecture: Hybrid (18 conv + 6 attention layers)
- Paper: LFM2 Technical Report
License
This model uses the LFM 1.0 License.
Citation
@article{liquidAI20268BA1B,
author = {Liquid AI},
title = {LFM2.5-8B-A1B: Personal Assistant On Your Laptop},
journal = {Liquid AI Blog},
year = {2026},
note = {www.liquid.ai/blog/lfm2-5-8b-a1b},
}
@article{liquidai2025lfm2,
title = {LFM2 Technical Report},
author = {Liquid AI},
journal = {arXiv preprint arXiv:2511.23404},
year = {2025}
}