Models · Hacker News ·
Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots
Cactus released Needle 2, a 45M-parameter agentic language model packaged as a 14MB binary for phones, wearables, smart-home devices, robots and microcontrollers. It runs sessions in 28MB of RAM and reports up to 500 tokens/sec on Raspberry Pi 5, with benchmark results comparable to some larger small models.