Tether releases QVAC BotPsy-0.8B, an open-weight 0.8B model for tool calls on personal devices

Tether AI Research announced the release of QVAC BotPsy-0.8B on November 11, 2026. The model is open-weight, has 0.8 billion parameters, and is designed for tool calls on personal devices.

Key details

• Parameters: 0.8 billion
• Base model: Alibaba's Qwen3.5-0.8B
• Licensing: open-weight
• Specialization: function calling and multi-step agentic workflows for local deployment
• Target tasks: calendars, messaging, files, device settings, web research and travel, among others

Performance claims

Tether AI Research reports that BotPsy-0.8B outperformed every tested model of up to 1.2 billion parameters across six agentic benchmarks. This result comes from Tether's own announcement. The benchmark names and individual scores are not given in the summary available at publication.

What this means

BotPsy-0.8B is a fine-tuned specialization of an existing small base model, Qwen3.5-0.8B, aimed at one job: reliably calling tools and chaining steps on the device itself. The comparison class Tether chose is models of up to 1.2 billion parameters, so its performance claim applies to that size range and has not been shown against larger models. Because the weights are open, the claim can be checked by anyone who runs the six benchmarks themselves.