Sao10K, who publishes as Sao, makes language-model merges and fine-tunes for creative writing and roleplay. The creator’s releases include Fimbulvetr, Stheno and Euryale.[1]
Start with L3.3-70B-Euryale-v2.3 →
Links
Website Hugging FaceBackground
Contributions
Euryale v2.3 builds on Llama 3.3 Instruct and is presented as a successor to Euryale v2.2. Its training configuration identifies datasets for roleplay, creative instruction and related dialogue, alongside LoRA adapter settings. The card supplies an instruction prompt format and sampling suggestions, while emphasizing that the system prompt affects results. These are the creator’s configuration and preferences, not a guarantee of quality for every conversation.[3]
Tools, writing & talks
Tools
Start here
L3.3-70B-Euryale-v2.3huggingface.coA Llama-based creative-dialogue model with a published training configuration and prompt guidance.
Start with the base-model, prompt-format and training-configuration sections. Running a 70B model needs a compatible inference service or suitable hardware. Compare outputs on your own creative-writing task and check the model licence before reuse.
Model creator[3]