anthracite-org/magnum-v4-12b

Model Overview

anthracite-org/magnum-v4-12b is a 12 billion parameter language model developed by anthracite-org, fine-tuned on the mistralai/Mistral-Nemo-Instruct-2407 base model. Its primary objective is to emulate the sophisticated prose quality found in Claude 3 Sonnet and Opus models, offering a 32768 token context window.

Key Capabilities

• Prose Quality Replication: Specifically trained to match the writing style and nuance of Claude 3 models.
• Extended Context: Features a 32768 token context length, enabling processing of longer inputs and generating more coherent, extended outputs.
• Instruction Following: Built upon an instruct-tuned base model, it is designed to follow complex instructions effectively.
• Training Data: Fine-tuned on a diverse set of datasets including anthracite-org/c2_logs_32k_llama3_qwen2_v1.2_no_system and various kalo-opus-instruct datasets, focusing on high-quality conversational and creative text.

Performance Metrics

Evaluations on the Open LLM Leaderboard show an average score of 19.95. Specific scores include 33.93 on IFEval (0-Shot), 30.50 on BBH (3-Shot), and 28.93 on MMLU-PRO (5-shot).

Good For

• Creative Writing: Generating high-quality, nuanced, and stylistically rich text.
• Advanced Conversational AI: Applications requiring sophisticated dialogue and detailed responses.
• Roleplay and Storytelling: Its focus on prose quality makes it well-suited for immersive narrative generation.