Ryex/Tower-Plus-9B-abliterated-hf-data

TEXT GENERATIONPricing:Input $0.431 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:16kPublished:Nov 18, 2025License:cc-by-nc-sa-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Ryex/Tower-Plus-9B-abliterated-hf-data is a 9 billion parameter language model derived from Unbabel's Tower-Plus-9B, specifically modified to reduce censorship while retaining core functionalities. This model is designed for general language tasks, with a particular focus on applications where less restrictive content generation is desired. It maintains a 16384-token context length, making it suitable for processing longer inputs and generating comprehensive outputs.

Loading preview...

Model Overview

Ryex/Tower-Plus-9B-abliterated-hf-data is a 9 billion parameter language model that represents an effort to reduce censorship in the original Unbabel/Tower-Plus-9B model, while aiming to preserve its foundational capabilities. The model is built upon the architecture of Tower Plus, which is noted for its multilingual processing, particularly in translation tasks as indicated by its original citation.

Key Characteristics

  • Censorship Reduction: The primary modification in this version is an attempt to "uncensor" the base Tower Plus model, offering more flexibility in content generation.
  • Parameter Count: It features 9 billion parameters, providing a balance between performance and computational requirements.
  • Context Length: The model supports a substantial context window of 16384 tokens, enabling it to handle extensive inputs and generate detailed responses.
  • Original Functionality: While modified for censorship, the model aims to maintain the core functionalities of the original Tower Plus, which includes strong multilingual capabilities, as suggested by its research paper.

Usage and Recommendations

This model is compatible with both VLLM and Hugging Face Transformers. For optimal performance, especially with its long context, the developers recommend using VLLM. Proper prompt formatting is crucial for effective interaction. The original Tower Plus model, from which this is derived, is cited in the paper "Tower+: Bridging Generality and Translation Specialization in Multilingual LLMs" by Ricardo Rei et al. (arXiv:2506.17080).

Finetunes
2 models