What are the hardware requirements for LFM2.5 8B A1B Heretic?

LFM2.5 8B A1B Heretic requires approximately 17.3 GB of VRAM and supports a context window of 125K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed LFM2.5 8B A1B Heretic and how large is it?

LFM2.5 8B A1B Heretic is developed by Dingdust, a model with 8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate LFM2.5 8B A1B Heretic?

LFM2.5 8B A1B Heretic is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

LFM2.5 8B A1B Heretic by Dingdust — VRAM 17.3GB, 125K context

Name: LFM2.5 8B A1B Heretic
Author: Dingdust

LFM2.5 8B A1B Heretic is an open-source language model by Dingdust. Features: 8b LLM, VRAM: 17.3GB, Context: 125K, License: other, LLM Explorer Score: 0.25.

Arxiv:2511.23404 Arxiv:2607.15232 Abliterated Ar Base model:finetune:liquidai/l... Base model:liquidai/lfm2.5-8b-... Conversational De Decensored Edge En Endpoints compatible Es Fr Heretic It Ja Ko Lfm2.5 Lfm2 moe Liquid Pt Region:us Safetensors Sharded Tensorflow Uncensored Zh

Model Card on HF 🤗: https://huggingface.co/Dingdust/LFM2.5-8B-A1B-heretic

LFM2.5 8B A1B Heretic Parameters and Internals

LLM Name	LFM2.5 8B A1B Heretic
Repository 🤗	https://huggingface.co/Dingdust/LFM2.5-8B-A1B-heretic
Base Model(s)	LiquidAI/LFM2.5-8B-A1B-Base LiquidAI/LFM2.5-8B-A1B-Base
Model Size	8b
Required VRAM	17.3 GB
Updated	2026-09-23
Maintainer	Dingdust
Model Type	lfm2_moe
Model Files	0.8 GB: 1-of-18 1.0 GB: 2-of-18 1.0 GB: 3-of-18 1.0 GB: 4-of-18 1.0 GB: 5-of-18 1.0 GB: 6-of-18 1.0 GB: 7-of-18 1.0 GB: 8-of-18 1.0 GB: 9-of-18 1.0 GB: 10-of-18 1.0 GB: 11-of-18 1.0 GB: 12-of-18 1.0 GB: 13-of-18 1.0 GB: 14-of-18 1.0 GB: 15-of-18 1.0 GB: 16-of-18 1.0 GB: 17-of-18 0.5 GB: 18-of-18
Supported Languages	en ar zh fr de ja ko es pt it
Model Architecture	Lfm2MoeForCausalLM
License	other
Context Length	128000
Model Max Length	128000
Transformers Version	5.14.1
Tokenizer Class	TokenizersBackend
Padding Token	<\|pad\|>
Vocabulary Size	128000

Rank the LFM2.5 8B A1B Heretic capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to LFM2.5 8B A1B Heretic

Best Alternatives	Context / RAM	Downloads	Likes
LFM2.5 8B A1B	125K / 16.9 GB	84755	670
LFM2.5 8B A1B Wolof CPT V3	125K / 17 GB	136	1
...inal ToolBench Full SFT 1Epoch	125K / 16.9 GB	8867	2
LFM2 8B A1B	125K / 16.7 GB	30570	370
LFM2.5 8B A1B Base	125K / 16.9 GB	4201	29
LFM2.5 8B A1B UltraData L2 49M	125K / 16.9 GB	293	1
LFM2.5 8B A1B Opus Distil	125K / 16.9 GB	2127	5
LFM2.5 8B A1B	125K / 16.9 GB	1240	13
...s Agentic Coder Abliterated V3	125K / 16.9 GB	12	1
...s Agentic Coder Abliterated V2	125K / 16.9 GB	0	4

Note: green Score (e.g. "73.2") means that the model is better than Dingdust/LFM2.5-8B-A1B-heretic.