Cydonia 24B V4.1 is a 24-billion-parameter text-to-text language model created by TheDrummer and fine-tuned from the Mistral Small 3.2 24B base. It continues TheDrummer's Cydonia series, building on earlier variants such as Cydonia-22B and Cydonia-24B-v2.x, and the published descriptions point to a deliberate focus on improving coherence and writing quality. The architecture is open-source, and the model operates purely on text input and text output, which makes it straightforward to drop into existing language pipelines that do not require multimodal handling.

The model is positioned primarily for uncensored creative writing, roleplay, and narrative-heavy chat, with an emphasis on preserving mood, nuance, and consistent characterization across long exchanges. Its roughly 131,072-token context window supports extended conversations and intricate storylines, and reviewers note strong prompt adherence and recall as practical strengths. Generation can be steered through sampling parameters such as temperature, frequency penalty, and presence penalty, giving writers fine-grained control over tone and variation. These traits make it a natural fit for fiction drafting, long-form dialogue systems, and hobbyist or enterprise assistant tasks where long-context behavior matters more than raw reasoning benchmarks.