Ateron/Sketch-Cydonia-24B-V1.2

TEXT GENERATIONConcurrency Cost:2Model Size:24BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Mar 11, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Ateron/Sketch-Cydonia-24B-V1.2 is a 24 billion parameter language model developed by Ateron, built upon the Sketch-CydoniaV1.1 base model and further refined using a merge method. This iteration, V1.2, focuses on reducing "slop and repetition" in its outputs, aiming for more concise and efficient text generation. It is configured with a 32768 token context length and utilizes a specific merging strategy with ConiCat MS 3.2 AntiRep to enhance output quality.

Loading preview...

Sketch-Cydonia-24B-V1.2 Overview

Ateron's Sketch-Cydonia-24B-V1.2 is a 24 billion parameter language model, representing an iterative refinement over its predecessor, Sketch-CydoniaV1.1. This version is specifically engineered to reduce "slop and repetition", aiming for more streamlined and efficient text generation. It leverages a sophisticated merging strategy, incorporating ConiCat MS 3.2 AntiRep with varying densities and weights, to achieve its improved output characteristics.

Key Capabilities

  • Reduced Repetition: Designed to minimize redundant phrases and information in generated text.
  • Enhanced Conciseness: Focuses on producing more direct and to-the-point responses.
  • Iterative Refinement: Built on a proven base model with targeted improvements for output quality.

Good For

  • Applications requiring concise and non-repetitive text outputs.
  • Scenarios where efficiency in language generation is critical.
  • Developers looking for a 24B model with a focus on output clarity and reduced verbosity.