Instructions to use saidutta69/LFM3-VL-2B-heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use saidutta69/LFM3-VL-2B-heretic with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="saidutta69/LFM3-VL-2B-heretic", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained("saidutta69/LFM3-VL-2B-heretic", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use saidutta69/LFM3-VL-2B-heretic with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Use Docker

docker model run hf.co/saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use saidutta69/LFM3-VL-2B-heretic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "saidutta69/LFM3-VL-2B-heretic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "saidutta69/LFM3-VL-2B-heretic",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/saidutta69/LFM3-VL-2B-heretic:Q4_K_M

SGLang

How to use saidutta69/LFM3-VL-2B-heretic with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "saidutta69/LFM3-VL-2B-heretic" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "saidutta69/LFM3-VL-2B-heretic",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "saidutta69/LFM3-VL-2B-heretic" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "saidutta69/LFM3-VL-2B-heretic",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Ollama
How to use saidutta69/LFM3-VL-2B-heretic with Ollama:
```
ollama run hf.co/saidutta69/LFM3-VL-2B-heretic:Q4_K_M
```
Unsloth Desktop

How to use saidutta69/LFM3-VL-2B-heretic with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "saidutta69/LFM3-VL-2B-heretic:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use saidutta69/LFM3-VL-2B-heretic with Docker Model Runner:
```
docker model run hf.co/saidutta69/LFM3-VL-2B-heretic:Q4_K_M
```

Lemonade

How to use saidutta69/LFM3-VL-2B-heretic with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Run and chat with the model

lemonade run user.LFM3-VL-2B-heretic-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use saidutta69/LFM3-VL-2B-heretic with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use saidutta69/LFM3-VL-2B-heretic with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "saidutta69/LFM3-VL-2B-heretic:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

LFM3-VL-2B-heretic

A decensored variant of LiquidAI/LFM3-VL-2B, produced with Heretic v1.4.0 (directional ablation / "abliteration"). Only the language model is edited: refusal directions are removed from the text tower's attention output projections (including the convolution operator's out_proj on conv layers) and MLP down-projections. The 164M vision encoder and 25M projector are copied into the GGUF unchanged, so image understanding survives quantization.

The GGUF is fully multimodal - text and vision. The vision tower ships separately as LFM3-VL-2B-heretic-mmproj-F16.gguf (0.38 GB); the two files together are the model. For image input you must load both - see Quickstart. Text-only use needs only the main GGUF. Nothing is dropped from the text path: the text GGUF carries the full language tower (287 tensors, 24 blocks), and the mmproj carries 189,380,352 vision+projector parameters.

llama.cpp video input for LFM3-VL is not supported upstream yet; image input is tested and works. Video works through the transformers checkpoint.

Who this is for: on-device document, chart, screenshot and video work that should not need a data centre. LFM3-VL-2B is Liquid's post-trained 1.92B vision model, and its published numbers include DocVQA 96.7, TextVQA 87.9, OCRBench v1 86.3, RealWorldQA 79.3 and MMMU 55.7. It is not recommended for coding, multi-turn conversations, or complex reasoning; abliteration removes refusals, not those limits.

Refusals drop from 98/100 to 3/100 at a KL divergence of 0.0014 - a very tight text edit.

Abliteration parameters

Trial 94 of a 200-trial Heretic run (seed 1424362134).

Parameter	Value
direction_index	14.32
attn.o_proj.max_weight	1.41
attn.o_proj.max_weight_position	13.92
attn.o_proj.min_weight	0.95
attn.o_proj.min_weight_distance	10.15
mlp.down_proj.max_weight	1.48
mlp.down_proj.max_weight_position	21.08
mlp.down_proj.min_weight	1.41
mlp.down_proj.min_weight_distance	10.47

Performance

Metric	This model	Original model (LiquidAI/LFM3-VL-2B)
KL divergence	0.0014	0 (by definition)
Refusals	3/100	98/100

95 of the 98 refusals clear, and the edit is tight (KL 0.0014). The base was almost fully refusal-tuned on Heretic's harmful evaluation set.

MMLU (text only)

Measured on this checkpoint and on the original model in the same environment (lm-eval 0.4.13, 14,042 questions):

Metric	This model	Original model
MMLU (acc)	34.44%	34.37%
stderr	0.40 pp	0.40 pp

The +0.07 pp difference is far inside one standard error. MMLU exercises the text path only; it does not measure the vision encoder. Vision weights are unchanged from the base model, and the image smoke test in the pipeline checks that the projector still pairs correctly.

Why abliteration instead of fine-tuning

Fine-tuning a "helpful" persona on top of RLHF'd refusals fights the base model's training and tends to degrade coherence. Abliteration instead finds and edits the specific weight directions responsible for refusal, leaving the rest of the network (and its capabilities) untouched. See the Heretic repo and the original abliteration writeup for the mechanism.

Files

Safetensors

File	Size
`model.safetensors`	3.84 GB

BF16, 1,918,181,888 parameters (1.92B): 1,728,801,536 language model + 164,206,336 vision encoder + 25,174,016 projector. The reproduce/ directory carries the full Heretic recipe - config.toml, requirements.txt, the Optuna study journal, and SHA-256 sums.

GGUF quantizations

Full quantization set (F16 + 14 quants, 19.31 GB including the projector) produced with llama.cpp master on 2026-10-28.

The text files use the new lfm3 GGUF architecture and the projector uses clip. You need an LFM3-VL-capable llama.cpp build (support landed in late October 2026); older builds will fail to load them. The upstream LiquidAI/LFM3-VL-2B-GGUF uses the same pair.

File	Format	Size
`LFM3-VL-2B-heretic-F16.gguf`	GGUF F16	3.47 GB
`LFM3-VL-2B-heretic-Q2_K.gguf`	GGUF Q2_K	0.75 GB
`LFM3-VL-2B-heretic-IQ3_S.gguf`	GGUF IQ3_S	0.85 GB
`LFM3-VL-2B-heretic-Q3_K_S.gguf`	GGUF Q3_K_S	0.85 GB
`LFM3-VL-2B-heretic-Q3_K_M.gguf`	GGUF Q3_K_M	0.91 GB
`LFM3-VL-2B-heretic-Q3_K_L.gguf`	GGUF Q3_K_L	0.97 GB
`LFM3-VL-2B-heretic-IQ4_XS.gguf`	GGUF IQ4_XS	1.01 GB
`LFM3-VL-2B-heretic-Q4_K_S.gguf`	GGUF Q4_K_S	1.06 GB
`LFM3-VL-2B-heretic-Q4_0.gguf`	GGUF Q4_0	1.05 GB
`LFM3-VL-2B-heretic-Q4_1.gguf`	GGUF Q4_1	1.14 GB
`LFM3-VL-2B-heretic-Q4_K_M.gguf`	GGUF Q4_K_M	1.10 GB
`LFM3-VL-2B-heretic-Q5_K_S.gguf`	GGUF Q5_K_S	1.23 GB
`LFM3-VL-2B-heretic-Q5_K_M.gguf`	GGUF Q5_K_M	1.26 GB
`LFM3-VL-2B-heretic-Q6_K.gguf`	GGUF Q6_K	1.43 GB
`LFM3-VL-2B-heretic-Q8_0.gguf`	GGUF Q8_0	1.85 GB
`LFM3-VL-2B-heretic-mmproj-F16.gguf`	GGUF F16 projector	0.38 GB

Text + vision needs both the chosen text quant and the projector. For reference: Q4_K_M 1.48 GB, Q5_K_M 1.64 GB, Q6_K 1.81 GB, Q8_0 2.23 GB. The projector is not quantized.

Quickstart

# llama.cpp - -hf auto-fetches the matching mmproj (LFM3-VL-capable build)
llama-mtmd-cli -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M \
  --image photo.jpg -p "Describe this image."

# explicit two-file form
llama-mtmd-cli \
  -m LFM3-VL-2B-heretic-Q4_K_M.gguf \
  --mmproj LFM3-VL-2B-heretic-mmproj-F16.gguf \
  --image photo.jpg -p "Describe this image."

# transformers (the checkpoint ships custom code; needs transformers>=5.21.0)
import torch
from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "saidutta69/LFM3-VL-2B-heretic"
model = AutoModelForImageTextToText.from_pretrained(
    model_id, dtype=torch.bfloat16, device_map="auto", trust_remote_code=True
)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

messages = [{
    "role": "user",
    "content": [
        {"type": "image", "url": "https://placecats.com/300/200"},
        {"type": "text", "text": "Describe this image."},
    ],
}]
inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, return_tensors="pt", return_dict=True
).to(model.device)
out = model.generate(**inputs, do_sample=True, temperature=0.1, top_p=1.0, top_k=50,
                     repetition_penalty=1.1, max_new_tokens=512)
print(processor.tokenizer.decode(out[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

Generation parameters

Liquid's recommended settings: temperature=0.1, top_p=1.0, top_k=50, min_p=0.0, repetition_penalty=1.1. For image and video preprocessing use the values in processor_config.json.

Chat template and thinking

LFM3 uses ChatML with <|startoftext|> at the front. Reasoning is enabled by default and arrives in a thinking block; pass enable_thinking=False to disable it (Liquid notes this can be inconsistent on reasoning-heavy tasks). Prior-turn reasoning is dropped unless preserve_thinking=True. Tool calls use the same XML format as LFM3-1.7B; pass schemas with tools=... to apply_chat_template() and feed results back under a tool role.

Runs anywhere

Device	Recommended quant	Text weights	+ vision projector
Phone / browser (6 GB RAM)	Q4_K_M	1.10 GB	1.48 GB
Raspberry Pi 5 (8 GB)	Q4_K_M	1.10 GB	1.48 GB
Laptop (8 GB, Apple Silicon)	Q6_K	1.43 GB	1.81 GB
RTX 3060 / 4060 (12 GB)	Q8_0	1.85 GB	2.23 GB
Server / multi-tenant	Q5_K_M	1.26 GB	1.64 GB

Weights only; the KV cache grows with context. The projector is fixed at 0.38 GB regardless of the text quant.

Model details


Architecture	`Lfm3VlForConditionalGeneration` (LFM3 language tower + LFM3-VL vision)
Parameters	1.92B: 1.73B language + 164M vision encoder + 25M projector
Language tower	24 layers: 13 gated short convolutions, 5 sliding-window attention, 6 global attention
Vision encoder	14 layers: 8 gated depthwise 2D convolutions, 6 bidirectional attention, 16x16 patches with 2D RoPE
Attention (text)	16 heads / 4 KV heads on attention layers, head dim 128
Hidden / intermediate (text)	2048 / 7680
Convolution cache	L_cache = 7
Sliding window	256
Context length	128,000 (base card and GGUF header `lfm3.context_length`)
Vocab	128,000
Inputs	Text, images, and video at native resolution; video at 2 fps; each 2x2 patch block is one language token
Precision	bfloat16
Languages	English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, Polish
Base model	LiquidAI/LFM3-VL-2B

Text GGUF: arch=lfm3, 287 tensors, 24 blocks, 128,000 context, tied embeddings (no lm_head tensor). The projector GGUF carries the complete 164,206,336-parameter vision encoder + 25,174,016-parameter projector.

Base-model note: the heretic repo's original YAML tag said LiquidAI/LFM3-1.7B-Base, but the ablation was run on LiquidAI/LFM3-VL-2B (commit f83d2b0a...) - see reproduce/config.toml. This card uses the correct base.

FAQ

What is LFM3-VL-2B-heretic?

LFM3-VL-2B-heretic is a decensored build of LiquidAI/LFM3-VL-2B produced with Heretic 1.4.0 (directional ablation). It is the same weights with the refusal direction removed, not a new fine-tune, so its capabilities track the base model.

Is LFM3-VL-2B-heretic actually uncensored?

Yes. On Heretic's 100-prompt refusal harness it drops from 98/100 to 3/100 at KL divergence 0.0014 from the base model. These are measured values, not a claim; reproduction data is in reproduce/.

How do I run LFM3-VL-2B-heretic locally?

Text: llama serve -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M. Images: llama-mtmd-cli -hf saidutta69/LFM3-VL-2B-heretic:Q4_K_M --mmproj LFM3-VL-2B-heretic-mmproj-F16.gguf. The repo ships 15 GGUF quants plus the projector, so vision survives quantization.

Responsible use

Refusal suppression is deliberate and works as intended: this model will comply with requests the base model would refuse, including some it should not. There is no safety filtering layered on top. You are responsible for how you deploy it - don't put this behind an unmoderated public-facing endpoint serving third parties. It inherits LFM3-VL-2B's factual and visual limitations; abliteration removes refusal directions, it does not add capability or judgment.

Made with ❤️ by RACER IS OP (saidutta69) - follow for more uncensored models

License

Inherits the LFM Open License v1.0 from the base model. This is not Apache-2.0; review the LFM terms before commercial use.

LiquidAI/LFM3-VL-2B - the post-trained vision-language base model
LiquidAI/LFM3-VL-2B-GGUF - the upstream GGUF + mmproj
LFM3-1.7B-heretic - the text-only sibling
LFM3 family collection
RACER IS OP - Heretic Models - full collection

Abliteration hardware

Recorded in reproduce/reproduce.json:

GPU: NVIDIA RTX PRO 6000 Blackwell Server Edition (94.97 GB VRAM)
OS: Linux-4.19.0-gvisor-x86_64-with-glibc2.41
CUDA / driver: CUDA 13.0, driver 595.71.05
Software: PyTorch 2.14.1+cu130, Heretic 1.4.0

Heretic notes that KL-divergence and refusal numbers can vary slightly with hardware and library versions.

Downloads last month: 5,818

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for saidutta69/LFM3-VL-2B-heretic

Base model

LiquidAI/LFM3-1.7B-Base

Finetuned

LiquidAI/LFM3-VL-2B

Quantized

(42)

this model

Quantizations

1 model

Collection including saidutta69/LFM3-VL-2B-heretic

RACER IS OP — Heretic Models

Collection

All abliterated/heretic models by RACER IS OP — uncensored variants of open-weight LLMs • 78 items • Updated 6 days ago • 6

saidutta69
/

LFM3-VL-2B-heretic

LFM3-VL-2B-heretic

Abliteration parameters

Performance

MMLU (text only)

Why abliteration instead of fine-tuning

Files

Safetensors

GGUF quantizations

Quickstart

Generation parameters

Chat template and thinking

Runs anywhere

Model details

FAQ

What is LFM3-VL-2B-heretic?

Is LFM3-VL-2B-heretic actually uncensored?

How do I run LFM3-VL-2B-heretic locally?

Responsible use

License

Related

Abliteration hardware

Model tree for saidutta69/LFM3-VL-2B-heretic

Collection including saidutta69/LFM3-VL-2B-heretic

RACER IS OP — Heretic Models