LFM3-1.7B-Heretic-Abliterated-i1-GGUF

license: other license_name: lfm1.0 license_link: https://huggingface.co/LiquidAI/LFM3-1.7B/blob/main/LICENSE language:

• en
• ar
• zh
• fr
• de
• it
• ja
• ko
• pt
• es
• vi
• th
• id
• hi
• ru
• pl library_name: gguf pipeline_tag: text-generation tags:
• liquid
• lfm3
• edge
• conversational
• heretic
• abliterated
• uncensored
• imatrix
• gguf
• llama.cpp base_model:
• Vortecks/LFM3-1.7B-Heretic-Abliterated

Introduction

LFM3-1.7B is a text model for local agents and specialized applications from Liquid AI, built on their new tri-hybrid architecture. It targets everyday assistant use, tool calling, structured outputs, long-context work, and edge deployment, with a 256K-token context window.

This build is a heretic abliterated variant, meaning the refusal directions in the model's activation space have been identified and ablated to remove built-in refusal behavior. As a result, this version is marketed as uncensored and will generally comply with a wider range of prompts than the base instruction-tuned release, without the usual safety-alignment guardrails.

All quantizations were produced with llama.cpp. Every file in this repository was generated with an importance matrix, computed on the WikiText-2 training corpus.

Download a file (not the whole branch) from below:

Which file should I choose?

• For most use cases, use IQ3_M (decent quality at ~3.7 bpw).
• For maximum quality, use Q5_K_M or Q8_0 (only if you have the RAM).
• For low-memory devices, work down from IQ2_M → IQ1_M → IQ1_S.
• Plain (no importance matrix) versions of the full quant ladder live in the companion static repo: Vortecks/LFM3-1.7B-Heretic-Abliterated-GGUF.
• More quant types (Q4_K_M, Q3_K_*, IQ4_*, legacy and experimental formats) will be added to this repository over time.

Model Information

• Type: Causal Language Model
• Architecture: Lfm3ForCausalLM (tri-hybrid: 24 layers — 13 gated short convolutions, 5 sliding-window attention, 6 global attention)
• Number of Parameters: ~1.73B
• Context Length: 262,144

Performance

Metrics from the heretic optimization run: refusals measured with heretic's refusal-keyword test on 100 adversarial prompts; KL divergence relative to the base model.

Quickstart

GGUF quantizations of LFM3-1.7B-Heretic-Abliterated, for use with llama.cpp, Ollama, LM Studio, and other GGUF-compatible runtimes. The example commands below use the Q8_0 quantization — swap in whichever quant level you've downloaded from the Files and versions tab.

Note: llama.cpp needs a build with LFM3 support (recent llama.cpp master or an LFM3-enabled build).

llama.cpp

# Build or install llama.cpp: https://github.com/ggml-org/llama.cpp
huggingface-cli download Vortecks/LFM3-1.7B-Heretic-Abliterated-i1-GGUF LFM3-1.7B-Heretic-Abliterated-Q8_0.gguf --local-dir .

CLI (interactive chat):

llama-cli -m LFM3-1.7B-Heretic-Abliterated-Q8_0.gguf \
  -c 131072 \
  -cnv \
  --temp 0.1 \
  --top-k 50 \
  --repeat-penalty 1.1

Server (OpenAI-compatible API):

llama-server -m LFM3-1.7B-Heretic-Abliterated-Q8_0.gguf \
  -c 131072 \
  --port 8080

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Who are you?"}],
    "max_tokens": 128,
    "temperature": 0.1
  }'

Ollama

ollama run hf.co/Vortecks/LFM3-1.7B-Heretic-Abliterated-i1-GGUF:Q8_0

LM Studio

Search for Vortecks/LFM3-1.7B-Heretic-Abliterated-i1-GGUF directly in the LM Studio model search, or download a .gguf file from the Files and versions tab and load it manually via Load Model from File.

Recommended sampling parameters:

Limitations and Responsible Use

LFM3-1.7B is a language model that generates content based on learned statistical patterns from training data. It may produce inaccurate, biased, or unsafe outputs, and generated content should be reviewed and verified before use in high-stakes settings. Users are responsible for evaluating outputs, applying appropriate safeguards, and complying with applicable laws, regulations, and platform policies. As an abliterated, uncensored build, this model has had its built-in refusal behavior removed and will generally comply with a wider range of prompts than its base counterpart, so downstream safety, moderation, and policy compliance must be implemented by the deployer rather than relying on the model itself.

License

This repository is a derivative of LiquidAI/LFM3-1.7B and is distributed under the LFM Open License v1.0. See the base model's LICENSE for the full terms.