richardyoung/Qwen2.5-14B-Instruct-heretic

TEXT GENERATIONConcurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 24, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The richardyoung/Qwen2.5-14B-Instruct-heretic is a 14.7 billion parameter instruction-tuned causal language model, based on the Qwen2.5 architecture developed by Qwen. This model is a decensored version of the original Qwen2.5-14B-Instruct, created using the Heretic v1.4.0 tool. It features a 32,768 token context length, extendable to 131,072 tokens with YaRN, and is optimized for reduced refusals compared to its base model, making it suitable for applications requiring less restrictive content generation.

Loading preview...

Overview

This model, richardyoung/Qwen2.5-14B-Instruct-heretic, is a 14.7 billion parameter instruction-tuned causal language model. It is a decensored variant of the original Qwen/Qwen2.5-14B-Instruct, processed using the Heretic v1.4.0 tool. The base Qwen2.5 series, developed by Qwen, features significant improvements in knowledge, coding, and mathematics, leveraging specialized expert models.

Key Differentiators

  • Decensored Output: Achieves a refusal rate of 11/100 compared to 98/100 for the original model, making it suitable for use cases requiring less content filtering.
  • Reproducible: The decensoring process is reproducible, with details provided in the reproduce/README.md.

Core Capabilities (inherited from Qwen2.5-14B-Instruct)

  • Enhanced Instruction Following: Improved ability to follow complex instructions and generate structured outputs like JSON.
  • Long Context Support: Supports a context length of 32,768 tokens, extendable up to 131,072 tokens using YaRN for processing extensive inputs.
  • Multilingual: Provides support for over 29 languages, including major global languages.
  • Robustness: More resilient to diverse system prompts, enhancing role-play and chatbot condition-setting.

When to Use This Model

This model is particularly well-suited for applications where a less restrictive content policy is desired, and the ability to generate responses that might be filtered by standard instruction-tuned models is beneficial. Its strong base capabilities in coding, mathematics, and long-text generation make it versatile for various demanding tasks.

Finetunes
1 models