HeraFox

Independent AI research lab.

Most of our models, experiments, and research systems have historically remained private.

Over the past several months, we have conducted substantial internal research across training, data, model behavior, and evaluation. Much of that work will remain private, but parts of it have directly contributed to our public Luna-S series.

Luna-S

Luna-S is HeraFox's small language model series for fast roleplay and character interaction.

It is not intended to compete with large general-purpose models. The focus is low-latency generation, lightweight deployment, and solid roleplay behavior within a small parameter budget.

The first generation, Luna-S1, is currently in training.

ETA: Q3 2026
Focus: Fast roleplay / character interaction
Family: Small language models
Status: Training

Luna-S1 is being developed around responsive dialogue, character consistency, conversational flow, and efficient inference.

Public variants will be released here when ready.

HeraFox
Most work stays private. Some of it becomes Luna.

Fox adapts. Whales implode