Overview of Luna-Ethos

Luna-Ethos is a 4 billion parameter conversational AI model developed by beyoru, specifically engineered for immersive roleplay (RP) and natural chatting. Unlike standard instruction-tuned models, Luna-Ethos prioritizes tone, personality, and narrative flow to create more engaging and human-like interactions.

Key Capabilities & Differentiators

• Enhanced Roleplay: Delivers responses that are expressive, character-driven, and rich in narrative, making conversations feel dynamic.
• Improved Reasoning: Beyond stylistic generation, it connects ideas coherently and provides structured explanations, balancing creativity with logical thought.
• Balanced Output: Designed to offer answers that are both compelling to read and thoughtfully reasoned, combining stylistic richness with deeper thinking.

Training & Limitations

Luna-Ethos's training includes Claude Opus 4.6 (maximum-thinking trajectories) and an RP LoRA merged from the Luna model, with SFT for reasoning and RL for RP style. It is primarily optimized for English, and multilingual performance is not guaranteed. Users should note that it may occasionally prioritize tone over precision and is not suitable for professional advice.