DarkKitsune/qwen3.5-9b-qworus

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 13, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

DarkKitsune/qwen3.5-9b-qworus is a 9 billion parameter language model created by DarkKitsune, formed by a DARE-TIES merge of empero-ai/Qwythos-9B-v2 and PeppX/Ornith-1.0-9B-Uncensored. This model combines Qwythos' stability and bug-fixing capabilities with Ornith's project organization and scalability. It is decensored and designed for testing, potentially inheriting long-context support up to 1M tokens from its base model.

Loading preview...

qwen3.5-9b-qworus Overview

DarkKitsune/qwen3.5-9b-qworus is a 9 billion parameter language model resulting from a 60/40 DARE-TIES merge of empero-ai/Qwythos-9B-v2 and PeppX/Ornith-1.0-9B-Uncensored. The primary goal of this merge is to combine the strengths of both base models: Qwythos's noted stability and proficiency in bug fixing based on stderr output, and Ornith's ability to facilitate well-organized and scalable project setups.

Key Characteristics

  • Merge Composition: A DARE-TIES merge, with 60% weight from Qwythos-9B-v2 and 40% from Ornith-1.0-9B-Uncensored.
  • Decensored: Inherits decensored characteristics from both base models.
  • Potential Long Context: May inherit 1M token context support from Qwythos, though this is untested.
  • Intended Use: This model is explicitly for testing purposes and is not recommended for serious production projects.

Technical Details

The merge was performed using mergekit with a dare_ties method, bfloat16 dtype, and int8_mask enabled. The normalize parameter was set to true.