yethdev/qwen3.5-9b-manumit-v1

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 7, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The yethdev/qwen3.5-9b-manumit-v1 is a 9 billion parameter language model based on the Qwen3.5-9B architecture, fine-tuned by yethdev. This model is specifically modified to reduce refusal rates, making it less prone to denying requests compared to its base model. It maintains a 32768 token context length and is optimized for use cases requiring a less restrictive response generation. Benchmarks indicate a significant reduction in refusal rate from 100% to 75% while improving MMLU scores.

Loading preview...

Model Overview

yethdev/qwen3.5-9b-manumit-v1 is a 9 billion parameter language model derived from the Qwen3.5-9B base model. Developed by yethdev, this version has undergone a process referred to as "abliteration" using the manumit v1 tool. The primary goal of this modification is to reduce the model's tendency to refuse certain requests, often termed as "safeguards."

Key Capabilities & Differentiators

  • Reduced Refusal Rate: The model significantly lowers the refusal rate from 100% (of the base Qwen3.5-9B) to 75%, making it more permissive in its responses.
  • Improved MMLU Performance: Alongside reduced refusals, the model demonstrates an improvement in MMLU (Massive Multitask Language Understanding) scores, increasing from 64.7% to 71.8% compared to its base model.
  • Base Model: Built upon the robust Qwen3.5-9B architecture, retaining its 32768 token context length.
  • Ongoing Development: The manumit tool used for this modification is actively being developed, suggesting potential future improvements.

When to Use This Model

This model is particularly suitable for applications where:

  • Less Restrictive Responses are Desired: Users need a model that is less likely to deny or filter requests based on built-in safeguards.
  • Specific Content Generation: Use cases that might be constrained by the default refusal mechanisms of standard LLMs.
  • Balancing Performance and Permissiveness: When a balance between general language understanding (as indicated by MMLU) and a lower refusal rate is critical.

Finetunes
2 models