criscarleo/Qwen2.5-Coder-14B-Instruct-abliterated

TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kPublished:Mar 27, 2026License:agpl-3.0Architecture:Transformer Open Weights Cold

criscarleo/Qwen2.5-Coder-14B-Instruct-abliterated is an experimental, instruction-tuned model processed using the Obliteratus methodology. This model is designed for code-related tasks, leveraging the Qwen2.5 architecture. It is provided in GGUF format for efficient inference with llama.cpp, making it suitable for local deployment and development workflows.

Loading preview...

Model Overview

This model, criscarleo/Qwen2.5-Coder-14B-Instruct-abliterated, is an experimental, instruction-tuned variant of the Qwen2.5 architecture. It has been processed using the Obliteratus methodology, a transformation technique developed by pliny-the-prompter. The fine-tuning and transformation process was executed via an official Jupyter Notebook provided within the Obliteratus Space.

Key Characteristics

  • Processing Methodology: Utilizes the Obliteratus method for fine-tuning and transformation.
  • Instruction-Tuned: Optimized for following instructions, likely enhancing its utility in interactive or task-oriented scenarios.
  • Inference Format: Provided in GGUF (v3) format, making it compatible with llama.cpp for efficient CPU-based inference.
  • Experimental Status: Marked as a "Work in Progress - Experimental version," indicating ongoing development and potential for further refinement.

Intended Use Cases

Given its instruction-tuned nature and the "Coder" designation in its name, this model is likely intended for:

  • Code Generation: Assisting with writing code snippets or functions.
  • Code Understanding: Explaining code, debugging, or refactoring suggestions.
  • Instruction Following: Executing specific coding-related commands or tasks.
  • Local Development: Its GGUF format makes it suitable for running on consumer hardware using llama.cpp.

Finetunes
1 models