Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback and have now incorporated suggestions to release Needle 2.

The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45M parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400–1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges 300–700 on sub-$200 phones. On tool call and mobile device use benchmarks, Needle 2 trades wins with closest small models like LFM2.5 230M and Apple Foundation Model, at 5x to 70x smaller.

Needle is based on Simple Attention Networks. Edge AI has lately meant Macs and PCs, but that is just 1.5 billion of over 21 billion connected IoT devices. Budget phones, Raspberry Pis, microcontrollers, wearables, small robots like Reachy Mini, and connected home devices are the target. Needle spends 7x to 85x fewer FLOPs per token than the smallest performant LLMs.

Every product has its own tool vocabulary and fine-tuning Needle helps it achieve frontier-level performance. Using the python package, Needle can be fine-tuned on a Mac/PC in minutes to a few hours. Combining Needle 2 with a private DeepSeek-v4-Flash deployment works well for enterprise tasks. Please use the playground to test and share thoughts.