N-Bot-Int/OpenElla-NovelWriter-Requiem-Ascended

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Jul 13, 2026License:agpl-3.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

N-Bot-Int/OpenElla-NovelWriter-Requiem-Ascended is an 8 billion parameter Llama 3.1-based language model developed by N-Bot-Int, specifically fine-tuned for enhanced roleplay capabilities. It addresses common issues like 'blarbing' and 'god-modding' found in previous NovelWriter versions, offering improved stability in prompt following and greater creative freedom. This model excels in character card sensitivity and diverse roleplay scenarios, making it suitable for interactive storytelling applications.

Loading preview...

Overview

N-Bot-Int/OpenElla-NovelWriter-Requiem-Ascended is an 8 billion parameter model developed by N-Bot-Int, representing their "best Llama model yet" for roleplay (RP). It is built upon a blend of previous OpenElla-NovelWriter versions (V1, V2, V3) and uses akjindal53244/Llama-3.1-Storm-8B as a stabilizing base. The model was fine-tuned using MERGEKIT on Kaggle's free tier.

Key Capabilities

  • Enhanced Roleplay (RP): Significantly improved RP capabilities compared to older NovelWriter generations.
  • Character Card Sensitivity: The model is sensitive to character cards, allowing for more consistent and in-character responses.
  • Creative Freedom: Offers a unique blend of creative freedom across various roleplay types.
  • Addresses "Blarbing" and "God-Modding": Solves issues where previous models would output entire chats without waiting for user prompts or exhibit "god-modding" behavior.
  • Prompt Following Stability: Demonstrates better stability in adhering to prompt instructions.
  • Uncensored Nature: The model is uncensored, requiring users to handle it with care and ethical considerations.
  • Activation Steering: Includes an obedience_vector.pt for activation steering at layers.23.mlp (strength 1.0–2.5) to improve character card compliance and suppress narrator takeover/OOC breaks.

Good For

  • Interactive Storytelling: Ideal for applications requiring dynamic and responsive roleplay.
  • Creative Writing: Suitable for generating diverse and creative narrative content.
  • Users Seeking Uncensored Models: For those who require an uncensored model for specific creative or research purposes, with the understanding of responsible and legal use.
  • Llama 3 Format Users: Best utilized with Llama 3 formatting for optimal performance.

Finetunes
1 models