Tether AI Research has released BotPsy, a 0.8-billion-parameter language model built for function calling and agentic workflows and designed to run on-device. The model is built on Alibaba's Qwen3.5-0.8B and targets simple to medium-difficulty tasks such as web search, travel and lifestyle planning, commerce and finance, device and vehicle control, file handling, calendar management and email. The company says it reaches state-of-the-art tool-use performance for its size, outperforming every model of up to 1.2 billion parameters that it tested. The weights are published on Hugging Face in safetensors format, with GGUF builds at bf16, Q8 and Q4 precision. The reported gains over the base model are large. On BFCL v4, the score rises from 23.4 to 44.0, on ACEBench from 30.1 to 58.7, and on τ-bench from 23.2 to 51.5. Against Qwen3.5-2B-Think, a model 2.5 times its size, BotPsy leads on five of six agentic benchmarks. The larger model edges ahead only on Claw-Eval, scoring 48.9 to BotPsy's 46.5. Tether also reports that instruction-following scores on IFEval rose from 59.7 to 90.8, and that BotPsy scored 86.1 on BFCL's hallucination tests, which check whether a model declines when no suitable tool exists. In its own automated red-teaming, Tether says the model is safer than leading models in its class. The company is open about the limits: the model still makes mistakes, is not intended for high-stakes use, becomes less reliable as instructions accumulate, and performs worse in lower-resource languages. All of these figures are company-reported. For Australian businesses and professionals, the practical impact is limited for now. BotPsy is a small model for low-stakes automation, not a replacement for frontier cloud models on complex work. Teams that run high volumes of routine tool calls could eventually reduce cloud API costs by keeping those tasks on local hardware. On-device processing also keeps data off third-party servers, which may ease obligations under the Privacy Act 1988 and the Australian Privacy Principles. Any deployment should still have a human review actions taken on a customer's behalf, and the model card's licence terms should be checked before commercial use. A developer or IT team could test the GGUF build on a non-sensitive internal workflow at little cost, but most other businesses can safely ignore it for now.