Core_24B_V.1 Open-Source Fusion Model - Retains Russian Processing Capability and Significantly Enhances Conversation Performance

Core 24B V.1

Developed by OddTheGreat

An upgraded fusion model based on Mistral 3.1 architecture, retaining Russian processing capabilities and enhancing dialogue performance

Large Language Model

Transformers

Supports Multiple Languages#Multilingual Role-Playing #Mistral 3.1 Architecture #Russian Optimization

Downloads 31

Release Time : 3/23/2025

Model Overview

This model is created by merging foundational models such as Gryphe/Pantheon-RP-1.8-24b-Small-3.1 and mrfakename/mistral-small-3.1-24b-instruct-2503-hf, focusing on role-playing and creative generation tasks.

Model Features

Multi-model Fusion

Combines Pantheon's expression imitation capabilities with the architectural advantages of Mistral 3.1

Bilingual Support

Fully retains Russian processing capabilities while supporting English

Dialogue Enhancement

Significantly improves dialogue performance while maintaining stability

Creative Generation

Excellent instruction-following ability and creative output

Model Capabilities

Text generation

Role-playing

Multi-turn dialogue

Creative writing

Bilingual processing

Use Cases

Entertainment

Role-playing Games

Generate coherent character dialogues and plots for role-playing games

Can produce noticeably different dialogue options

Creative Writing

Story Creation

Assist writers in creative writing and story ideation

Excellent narrative performance

Multilingual Applications

Russian Content Generation

Generate Russian text content

No anomalies in both full Russian cards and partial translation cards

🚀 [Merged Pretrained Language Model]

This project merges multiple pretrained language models to enhance dialogue capabilities while maintaining stability and performance in multiple languages.

🚀 Quick Start

The model is a merge of several pretrained language models. With the recent release of the new Mistral, which is slightly better than its predecessor, the goal was to update Apparatus. After testing the new Pantheon and being impressed by its ability to mimic human expressions and mannerisms, the idea of this new merge was born.

✨ Features

Model Merge: Combines multiple base models (Gryphe/Pantheon-RP-1.8-24b-Small-3.1, mrfakename/mistral-small-3.1-24b-instruct-2503-hf, OddTheGreat/Apparatus_24B) to create a new language model.
Enhanced Capabilities: Transfers Apparatus to the new Mistral 3.1, enhancing dialogue capabilities while preserving stability and Russian language performance.
Multilingual Support: Supports both English and Russian.

📦 Installation

No specific installation steps were provided in the original document, so this section is skipped.

💻 Usage Examples

No code examples were provided in the original document, so this section is skipped.

📚 Documentation

Model Details

Property	Details
Base Models	Gryphe/Pantheon-RP-1.8-24b-Small-3.1, mrfakename/mistral-small-3.1-24b-instruct-2503-hf, OddTheGreat/Apparatus_24B
Library Name	transformers
Tags	mergekit, merge, roleplay, creative
Languages Supported	English, Russian

Performance

The model has been tested on 250 answers. It shows good performance in narration and dialogues. The swipes make a difference in the output. The Russian performance remains intact, with no issues on full Russian cards and partly translated ones.

Known Issues

Qvink Memory: Enabling Qvink Memory seems to break something, causing replies to become much shorter.
Format Recommendation: It is better to use the prebuilt V7 format in ST, but ChatML also works fine. The recommended settings are T1.01 XTC 0.1 0.1.

Feedback Request

The author is not sure if this model is better than Apparatus and welcomes user feedback.

Future Plans

The model currently does not include vision capabilities. The author plans to figure out how to merge it in and will update the model accordingly.

🔧 Technical Details

No technical implementation details were provided in the original document, so this section is skipped.

📄 License

No license information was provided in the original document, so this section is skipped.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご