Notes

Lucy is a compact but capable 1.7B model focused on agentic web search and lightweight browsing, developed by Menlo Research.

• Temp: The official recommendation is a higher temp (0.7), which makes sense for its creative/search-agent role. Start there, but lower it to ~0.5 if you need more deterministic outputs.

Description

Lucy (1.7B) — A compact, agentic model for web search and lightweight browsing. Built on Qwen3-1.7B and optimized with machine-generated task vectors and pure RL, it’s designed to run efficiently on mobile/CPU devices.

Ideal for: - Agentic web search (using tools like Serper) - Lightweight web browsing and data extraction - On-device AI for mobile or low-resource setups

References

HuggingFace