Model Overview

anthracite-org/magnum-v2-32b is a 32.5 billion parameter language model, built upon the Qwen1.5 32B architecture, with a substantial context length of 32768 tokens. Developed by Anthracite, this model is the third iteration in a series focused on emulating the sophisticated prose quality found in Claude 3 models, specifically Sonnet and Opus.

Key Capabilities

• Claude 3 Prose Replication: Fine-tuned to achieve a similar writing style and quality as Claude 3 Sonnet and Opus.
• Instruction Following: Instruct-tuned using ChatML formatting, enabling effective conversational interactions.
• Extensive Training Data: Leverages a diverse set of filtered datasets, including Stheno, Claude 3.5 single-turn conversations, PhiloGlanSharegpt, Magpie-Reasoning-Medium-Subset, Opus_Instruct_25k, Opus_WritingStruct, and a subset of Sonnet3.5-SlimOrcaDedupCleaned.
• Robust Training: Underwent full-parameter fine-tuning over 2 epochs using 8 NVIDIA H100 Tensor Core GPUs.

Good For

• Applications requiring high-quality, human-like text generation.
• Advanced conversational AI and chatbot development.
• Creative writing and content generation where prose style is critical.
• Use cases benefiting from a large context window for complex interactions.