Tether AI Research has released QVAC BotPsy-0.8B, an open-weight artificial intelligence model with only 800 million parameters that is designed to execute multi-step tool calls directly on personal devices. Announced on November 11, 2026, the model targets everyday phone and assistant tasks such as scheduling, messaging, troubleshooting, and travel arrangements.
In Tether’s testing across six agentic benchmarks—BFCL v4, ACEBench, τ-bench, Claw-Eval, BrowseComp-Plus, and ToolSandbox—BotPsy outperformed every evaluated model with 1.2 billion parameters or fewer. It nearly doubled the scores of its base architecture, Qwen3.5-0.8B-Think: BFCL v4 rose from 23.4 to 44.0, ACEBench from 30.1 to 58.7, τ-bench from 23.2 to 51.5, and ToolSandbox from 30.5 to 64.6. Tether said the model also beat Alibaba’s Qwen3.5-2B-Think despite being 2.5 times smaller, posting 46.1 points on BFCL multi-turn tool calling versus 23.8 for the larger model and 7.5 for its base version.
BotPsy is designed to recognize when tools should be used, when multiple steps are needed, and when no available tool fits a request. In one ACEBench example, the model activated Wi-Fi and cleared inbox space before sending a message that had failed, rather than repeatedly retrying the same action. It also scored 86.1 on BFCL hallucination tests, the highest among models Tether evaluated, and reached 82.5 on τ-bench telecom tasks.
The release extends Tether’s QVAC ecosystem, which includes VisionPsy, TranscribePsy, MedPsy, and TranslatePsy. CEO Paolo Ardoino framed the effort as a move toward local AI that gives users more control over personal data and reduces reliance on centralized servers. Tether’s longer-term vision is for BotPsy to power a personal QVAC companion, with smartphones handling daily interactions and home computers or trusted peer-to-peer nodes processing more demanding requests, potentially routing searches through encrypted P2P connections rather than corporate cloud infrastructure.
