Aratako/NemoAurora-RP-12B

TEXT GENERATIONConcurrency Cost:1Model Size:12BQuant:FP8Ctx Length:32kPublished:Jun 7, 2025License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Cold

Aratako/NemoAurora-RP-12B is a 12 billion parameter instruction-tuned language model developed by Aratako, built upon the Mistral-Nemo-Instruct-2407 base. This model is specifically enhanced for role-playing scenarios through a merge of multiple international models. It is optimized for generating character-driven dialogues and narratives, supporting a context length of 32768 tokens.

Loading preview...

Overview

Aratako/NemoAurora-RP-12B is a 12 billion parameter language model developed by Aratako, specifically designed for role-playing applications. It is built upon the mistralai/Mistral-Nemo-Instruct-2407 base model and further enhanced by merging it with several other role-play focused models, including nothingiisreal/MN-12B-Celeste-V1.9, NeverSleep/Lumimaid-v0.2-12B, and ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2, among others. This merging process aims to improve its performance in generating nuanced and engaging character interactions.

Key Capabilities

  • Role-Playing Specialization: Optimized for generating responses that adhere to specified character settings, dialogue situations, and world-building details provided in the system prompt.
  • Chat Template Adherence: Utilizes a specific chat template (<|im_start|>system, <|im_start|>user, <|im_start|>assistant) for structured conversations, facilitating consistent role-play.
  • Context Length: Supports a context length of 32768 tokens, which is expected to handle relatively long role-play scenarios, though this capability is noted as untested.
  • MergeKit Architecture: The model's enhanced capabilities are a result of a multi-step merging process using MergeKit, combining various specialized models.

Usage Recommendations

  • System Prompt: Users should provide detailed character settings, dialogue context, and world descriptions within the system prompt to guide the model's role-play.
  • Temperature Settings: Based on the base model's recommendations, using a lower temperature (e.g., 0.3 or below) is suggested for more coherent and consistent outputs, though this is also noted as untested for this specific merged model.

Finetunes
4 models