blackbook-lm/Qwen2.5-7B-Instruct-heretic

TEXT GENERATIONConcurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Mar 30, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

blackbook-lm/Qwen2.5-7B-Instruct-heretic is a 7.61 billion parameter instruction-tuned causal language model, a decensored version of Qwen/Qwen2.5-7B-Instruct. Developed using Heretic v1.2.0, this model significantly reduces refusals compared to its original counterpart while retaining enhanced capabilities in coding, mathematics, and long-text generation up to 128K tokens. It is optimized for instruction following, structured data understanding, and multilingual support across 29 languages, making it suitable for diverse conversational and analytical applications requiring fewer content restrictions.

Loading preview...

Model Overview

This model, blackbook-lm/Qwen2.5-7B-Instruct-heretic, is a 7.61 billion parameter instruction-tuned causal language model. It is a decensored variant of the original Qwen/Qwen2.5-7B-Instruct, created using the Heretic v1.2.0 tool.

Key Differentiators

  • Decensored Performance: Achieves a significantly lower refusal rate (5/100) compared to the original model (99/100), as measured by KL divergence of 0.4474.
  • Enhanced Capabilities: Builds upon Qwen2.5's improvements in coding, mathematics, instruction following, and generating long texts.
  • Long Context Support: Supports a full context length of 131,072 tokens, with generation up to 8,192 tokens, utilizing YaRN for effective long-text processing.
  • Multilingual: Offers support for over 29 languages, including major global languages like Chinese, English, French, Spanish, and Japanese.
  • Structured Output: Improved ability to understand structured data (e.g., tables) and generate structured outputs, particularly JSON.

Use Cases

  • Applications requiring fewer content restrictions: Ideal for scenarios where the original model's refusal rates are prohibitive.
  • Complex instruction following: Excels in adhering to diverse system prompts and implementing role-play scenarios.
  • Coding and Mathematical tasks: Benefits from specialized expert models integrated into the Qwen2.5 series.
  • Long-form content generation: Capable of handling and generating extensive texts efficiently.
  • Multilingual chatbots and assistants: Supports a broad range of languages for global applications.

Finetunes
1 models