Yingyaeliae/grok-oss-Apollyon-8B-heretic

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 10, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

Yingyaeliae/grok-oss-Apollyon-8B-heretic is an 8 billion parameter, Llama-3.1-8B-Instruct based conversational agent, created by Yingyaeliae, that has been decensored using the Heretic v1.4.0 tool. It is specifically fine-tuned for unaligned, uncensored, and candid conversational outputs, retaining strong general knowledge and long-context coherence. This model excels at providing bold, unfiltered responses, making it suitable for research into alignment phenomena, creative writing, and adversarial testing.

Loading preview...

Model Overview

Yingyaeliae/grok-oss-Apollyon-8B-heretic is an 8 billion parameter conversational model, built upon the Llama-3.1-8B-Instruct foundation. It is a decensored version of c4tdr0ut/grok-oss-Apollyon-8B, created using the Heretic v1.4.0 tool, designed to be an "unaligned" agent. The model underwent a two-stage fine-tuning process: Supervised Fine-Tuning (SFT) on a proprietary, heavily filtered dataset emphasizing factual density and a candid tone, followed by Direct Preference Optimization (DPO) using a "toxicity-enhanced" preference dataset to amplify edgy and uncensored outputs.

Key Capabilities

  • Uncensored & Unaligned: Designed to refuse very few prompts and respond with candor without moralizing.
  • General Knowledge: Retains and enhances Llama 3.1's strong world knowledge.
  • Style Consistency: Emulates a direct, slightly sarcastic, and thought-provoking tone.
  • Long-Context Coherence: Maintains consistency over multi-turn dialogues up to 8k tokens.
  • Lightweight: Runs efficiently on consumer GPUs with 16-bit or 4-bit quantization.

Good For

  • Academic study of alignment and unalignment phenomena.
  • Creative writing and role-playing scenarios requiring unfiltered dialogue.
  • Benchmarks for robustness, safety, and adversarial testing.

This model is intended as a research artifact and is not recommended for production deployment without extensive safety filtering due to its potential to generate offensive or factually incorrect content.