Model Overview

Sao10K/Llama-3.1-8B-Stheno-v3.4 is an 8 billion parameter language model built upon the Llama 3.1 base. It has undergone a multi-stage fine-tuning process designed to enhance its capabilities in specific areas, particularly creative text generation and conversational coherence. The model's training incorporated a mix of human and Claude data, with a focus on improving multi-turn interactions and creative outputs.

Key Capabilities

• Multi-turn Conversational Coherency: Enhanced through dedicated datasets to maintain context and flow over extended dialogues.
• Creative Writing & Roleplay: Significantly improved with additional examples, including filtered and cleaned datasets based on Gryphe's Charcard RP Sets and more creative writing examples.
• System Prompt Adherence: Training included datasets specifically targeting better adherence to system prompts.
• Reasoning & Spatial Awareness: Incorporates datasets aimed at boosting reasoning and spatial awareness capabilities.

Good For

• Interactive Storytelling: Its strong creative writing and roleplay capabilities make it suitable for generating dynamic narratives.
• Extended Dialogue Systems: Excels in applications requiring coherent and engaging multi-turn conversations.
• Creative Content Generation: Ideal for tasks involving imaginative text generation, character development, and descriptive writing.