Cactus Compute released Needle 2, a 14MB function-calling LLM that runs entirely on a Raspberry Pi 5's CPU and selected the correct tool for the prompt "Turn the LED on" in 78 milliseconds. The model,…
FamiliaSync replaced its hand-written intent router with a small on-device classifier for its offline-first family organizer assistant, training 105 intents on roughly 2,400 example utterances with 70…
A developer's field notes detail the implementation of on-device tool calling in a fitness app using the Needle 2 model, a 45M-parameter, 2-bit quantized model that runs entirely on CPU with about 28 …
Cactus Compute released Needle 2, a 14 MB agentic model designed for tool calling, device control, and structured data extraction on edge devices. A developer's hands-on testing revealed several bugs,…
A developer built react-native-needle, a React Native binding for the Needle 2 on-device LLM, and found that while it excels at tool-calling tasks with sub-300ms response times, it can fabricate numbe…
Cactus Compute released Needle 2, an open 45M-parameter tool-calling model that ships as a 14MB binary and runs a full session in about 28MB of RAM, targeting devices with no GPU or NPU. The model ach…
Cactus Compute's Needle 2 model fits a functional LLM into just 14MB, using Simple Attention Networks to reduce per-token computation to 70 MFLOPs, compared to 164 MFLOPs for a standard transformer of…
Cactus Compute, a Y Combinator Summer 2025 startup, released Needle 2, a 14MB language model for tool calling on low-memory devices, capable of running on phones, wearables, home devices, robots, and …
Cactus released Needle 2, a 14MB agentic LLM for tool call, device use, and structured extraction, running 45 million parameters at 2-bit compression in 28MB of RAM. It achieves 500 tokens/sec on a Ra…