Tether introduces BotPsy, a small open-weight model for local tool use
Tether AI Research introduced BotPsy on November 11, a 0.8-billion-parameter model based on Qwen3.5-0.8B and specialized for tool calling. Its team reports stronger results than its base model across agent benchmarks, including a BFCL v4 score of 44.0 versus 23.4. These are developer-run comparisons, not independent validation. Team announcement
The public repository lists a 1.71 GB safetensors weight file, supporting current availability despite future-tense release wording in the announcement. The model carries Apache-2.0 licensing. Files, license
The model card says training code and training data are not included. English is the recommended language, and the evaluated configuration enables thinking mode. It also identifies weak performance in deep research, web search and memory tasks, and warns against high-stakes applications involving money, health or personal information. Model card
The useful distinction is between a compact tool-selection model and a ready-to-trust autonomous assistant. Local execution can give developers more control over deployment, but it does not remove the need to constrain tool permissions and validate actions. No inference or device-performance tests were performed for this report.
Sources and citations
Revision history
Published .