BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model

Tether AI Research introduces BotPsy, a 0.8B-parameter language model designed for on-device personal assistant tasks and agentic workflows. Built on top of Qwen3.5-0.8B, the model specializes in function calling to perform a variety of simple tasks for users while maintaining data privacy. BotPsy achieves state-of-the-art performance for its size, outperforming its base model and even larger models like Qwen3.5-2B-Think on several tool-use and agentic benchmarks. The model is intended for secure, self-hosted workflows that can run efficiently on local devices without relying on third-party cloud LLMs.