saidutta69/DeepSeek-R1-Distill-Qwen-14B-heretic

TEXT GENERATIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 21, 2026License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

The saidutta69/DeepSeek-R1-Distill-Qwen-14B-heretic is a 14.8 billion parameter language model, derived from deepseek-ai/DeepSeek-R1-Distill-Qwen-14B. This model has been decensored using the Heretic v1.4.0 abliteration method, which suppresses refusal behavior via targeted weight edits rather than fine-tuning. It retains the base model's knowledge and instruction-following capabilities, making it suitable for developers seeking a powerful reasoning model without built-in refusal guardrails.

Loading preview...

Overview

This model, created by saidutta69, is a 14.8 billion parameter variant of the DeepSeek-R1-Distill-Qwen-14B architecture. Its primary distinction is the removal of refusal behaviors through a process called "abliteration" using Heretic v1.4.0. Unlike traditional fine-tuning, abliteration directly edits specific weights responsible for refusal, preserving the base model's core knowledge and instruction-following abilities.

Key Capabilities

  • Decensored Output: Suppresses refusal behavior, allowing the model to answer requests that the base model would typically decline.
  • Retained Base Model Intelligence: Maintains the original DeepSeek-R1-Distill-Qwen-14B's reasoning and instruction-following capabilities.
  • Efficient Deployment: Available in various GGUF quantizations (Q4_K_M, Q5_K_M, Q6_K, Q8_0) to run on consumer GPUs with 8GB to 24GB VRAM, as well as CPU-only or Apple Silicon.
  • Broad Compatibility: Supports llama.cpp, transformers, Ollama, LM Studio, Jan, vLLM, and SGLang.

Good For

  • Developers who require a powerful reasoning model without built-in refusal mechanisms.
  • Use cases where direct answers are preferred over guarded responses.
  • Experimentation with models that have had their safety guardrails intentionally removed for specific research or application needs.

Note: This model is intentionally designed to comply with requests the base model would refuse, including potentially unsafe ones. Users are responsible for its deployment and usage.

Finetunes
1 models