Open Source
Tether AI launches BotPsy, a sub-billion agentic model
Hugging Face Bloghigh confidence
Why it matters
- →Enables private, on-device agentic workflows without cloud dependency.
- →Sets a new performance bar for sub-billion parameter models in tool-use tasks.
- →Provides open weights for developers to build lightweight autonomous agents.
Tether AI Research has released BotPsy, an open-weight agentic language model with 0.8 billion parameters, designed specifically for on-device personal assistant tasks. Built upon the Qwen3.5-0.8B foundation, the model undergoes a four-stage post-training process including supervised fine-tuning and reinforcement learning to optimize tool-use capabilities. Benchmarks indicate it outperforms models up to 1.2B parameters in agentic performance metrics like BFCL v4 and ACEBench, while also demonstrating improved instruction following and reduced reasoning loops compared to its base model. The weights are available under the Apache 2.0 license in multiple quantizations, including GGUF formats compatible with llama.cpp.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More from Hugging Face Blog
© Hugging Face BlogModels & Labscoding
IBM releases LLMON markup to fix prompt injection
IBM is tackling the brittleness of small language models by introducing LLMON, a structured markup language that explicitly separates instructions from data. By using reserved special tokens instead of ambiguous characters like JSON brackets, these new Granite 4.1 preview models drastically reduce indirect prompt injection attacks while maintaining standard performance. This approach forces the model to bind specific constraints to specific data payloads, eliminating the confusion that often leads to hallucinations or security breaches in long contexts.
Hugging Face Blog
© Hugging Face Blog
