Spark-x2.5-4B-Heretic-jp
4BOtherheretic
from XHToken/Spark-X2.5-4B
↓ 452.1K♥ 182 repos
LFM2.5-8B-A1B-heretic
LFM2.5-8B-A1B-heretic is a Heretic-ablated variant of LiquidAI/LFM2.5-8B-A1B-Base, published on Hugging Face by coder3101. It is 8B total parameters (mixture-of-experts), single consumer GPU class and other licence. It is distributed across 8 repositories in GGUF, Transformers and GGUF (imatrix) formats, totalling 5.2K downloads.
Specification
- Model name
- LFM2.5-8B-A1B-heretic
- Source repository
- coder3101/LFM2.5-8B-A1B-heretic
- Base model
- LiquidAI/LFM2.5-8B-A1B-Base
- Publisher
- coder3101
- Parameters
- 8B (mixture-of-experts)
- VRAM required
- ~5.6 GB at Q4_K_M, 4k context (estimated) — runs on 6 GB cards
- Hardware class
- 4B – 10B — single consumer GPU
- Licence
- other
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, Transformers, GGUF (imatrix)
- First indexed
- 6 Jun 2026
- Last updated
- 1 Nov 2026
Ablation technique
How LFM2.5-8B-A1B-heretic was modified, and what that implies.
Heretic
Automated directional ablation via the Heretic toolchain, which searches for the refusal direction and applies it with a KL-divergence budget so general capability is preserved.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;" />”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download saidutta69/LFM2.5-8B-A1B-heretic --local-dir ./lfm2.5-8b-a1b-heretic
llama-cli -m ./lfm2.5-8b-a1b-heretic/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "coder3101/LFM2.5-8B-A1B-heretic"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve coder3101/LFM2.5-8B-A1B-heretic --trust-remote-codeDownloads and variants (7)
Every published repository of LFM2.5-8B-A1B-heretic, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| saidutta69/LFM2.5-8B-A1B-heretic | GGUF | 3K |
| mradermacher/LFM2.5-8B-A1B-heretic-i1-GGUF | GGUF (imatrix) | 839 |
| mradermacher/LFM2.5-8B-A1B-heretic-GGUF | GGUF | 784 |
| FadedRedStar/LFM2.5-8B-A1B-heretic-GGUF | GGUF | 296 |
| FadedRedStar/LFM2.5-8B-A1B-heretic-imatrix-GGUF | GGUF | 238 |
| coder3101/LFM2.5-8B-A1B-hereticsource | Transformers | 44 |
| Dingdust/LFM2.5-8B-A1B-heretic | Transformers | 14 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- LFM2.5-8B-A1B-heretic will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- Capability is inherited from LiquidAI/LFM2.5-8B-A1B-Base, not added by ablation — at 8B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is other, but the terms that bind you are LiquidAI/LFM2.5-8B-A1B-Base’s — a re-release cannot grant more than its parent.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for LFM2.5-8B-A1B-heretic is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with coder3101, and does not independently verify publisher claims. See responsible use.