BotPsy: A Tiny 0.8B Personal Assistant Model That Beats Larger AI
Tether AI Research introduces BotPsy, a 0.8B-parameter model designed for on-device agentic workflows. Built on Qwen3.5-0.8B, it achieves state-of-the-art tool use for its size, outperforming many larger models while keeping user data private.
Original image temporarily unavailable
BotPsy was developed by Tether AI Research to address the high costs, privacy risks, and operational vulnerabilities of relying exclusively on cloud-based frontier LLMs. By focusing on a sub-billion-parameter size, the model is optimized for efficient, secure on-device deployment.
Training Methodology
The model undergoes a four-stage post-training process. This begins with supervised fine-tuning on tool-use conversations, followed by multi-teacher on-policy distillation using specialized 4B teachers. It concludes with two rounds of reinforcement learning using rule-based and verification-driven rewards to master tool usage and general tasks.
Performance and Efficiency
Despite having only 0.8B parameters, BotPsy roughly doubles the agentic capability of its base model. It outperforms several larger models, including the 2.5× larger Qwen3.5-2B-Think, on multiple tool-use and multi-turn benchmarks while using fewer reasoning tokens.
View selection rationale
Importance 79/100
About this brief
- Content type
- News summary
- Source type
- First-party / official source
- Original source
- Hugging Face ↗
- Source published
- Process
- AI-assisted discovery, translation, summarisation, classification and scoring; standards and corrections are maintained by the editorial desk.
- Verification
- The original URL and publication metadata are retained. This brief is not independently fact-checked unless explicitly stated.
- Published by
- Polyemergence Editorial Desk · PolyEmergence Pte Ltd.
- Conflict disclosure
- No financial or commercial conflict declared.