Anthracite ORGmodel publisher
Anthracite ORG
This is a Model Atlas entity, not a ranked standalone lab. A community or pseudonymous model publisher, not a verified standalone lab.
Anthracite illustrates how small communities can shape model behavior through data curation and fine-tuning without training a base model from scratch.
Profile
Overview
Anthracite is an open-model community that publishes conversational fine-tunes and datasets on Hugging Face. Its best-known line is Magnum, a family tuned for long-form roleplay and expressive instruction following.
The group builds on open base models and documents training datasets, merge recipes, and compute contributors. It operates more like a distributed research collective than a conventional incorporated laboratory.
Notable contributions
- 01Magnum roleplay fine-tunesThe project assembled large conversational datasets and full-parameter tuning runs aimed at expressive long-form dialogue.