Tether AI Research has introduced BotPsy, a 0.8-billion-parameter model built for function calling and agentic workflows. According to the source, the model is built on Qwen3.5-0.8B and is designed for on-device deployment.
The source describes BotPsy as a personal assistant that uses tools to handle simple tasks in domains including web search and research, travel and lifestyle, commerce and finance, business and customer operations, device and vehicle control, files and documents, calendar, and email and messaging. The source says the model is best suited to tasks of low to medium difficulty.
The source reports that BotPsy outperforms every model of up to 1.2 billion parameters that Tether tested, and calls this state-of-the-art tool-use and agentic performance for its size. The source also says the model makes mistakes and is not intended for high-stakes use.
Training has four sequential stages: supervised fine-tuning, multi-teacher on-policy distillation, and two rounds of reinforcement learning. The source says the distillation stage uses three Qwen3.5-4B teacher models, each specialized in tool use, general reasoning, or instruction following.
On the tool-use benchmarks listed in the source, BotPsy scores 44.0