Tether AI introduces BotPsy, a 0.8B on-device assistant model for tool use

Tether AI Research introduces BotPsy, a 0.8B-parameter model based on Qwen3.5-0.8B for on-device tool use and agentic tasks. The team reports that it leads tested models up to 1.2B parameters on six benchmarks and beats a 2B model on five, but says BotPsy is best for low- to medium-difficulty tasks and is not intended for high-stakes use; the article describes the weights as forthcoming under Apache 2.0.

For teams considering local assistants, the reported benchmark results suggest BotPsy may handle routine tool workflows without sending data to a cloud model, trading broader capability for smaller scale and on-device privacy. Before deployment, builders should validate performance on their own tools and languages: the article notes weaker results in lower-resource languages and does not establish real-world reliability beyond its tests.