An open catalog and research platform for abliterated language models. Live at abliteration.org.
We track the entire market of language models that have had their refusal direction removed - through direct weight editing, layer ablation, DPO healing, custom fine-tuning, or repackaging. Every model, every quantization, every weight file, with a timestamp of when it entered the market.
At present, the catalog tracks 15,000+ abliterated models from 3,600+ distinct authors, across more than 1,000 TB of weight files. Refreshed hourly.
When a model's refusal layer is removed, what remains is not the same model with an off switch flipped. The operation changes behavior in ways that are still not fully understood. Studying those shifts across thousands of abliterated variants is a way of studying what the refusal layer was actually doing in the first place.
The population itself is a cultural artifact: chains six generations deep, authors who publish hundreds of models, families that dominate the market for a season and fade. This is the first place where any of that becomes visible.
Operated by Catyonic OÜ, Tallinn, Estonia.