Cactus Compute releases tiny 45M-param Needle 2 model for devices

Cactus Compute has released Needle 2, a compact 45-million-parameter open-source model designed for tool calling, device integration, and structured data extraction. This model is notable for its small size, requiring only 28MB of RAM and fitting into a 14MB binary, making it capable of running on microcontrollers and various consumer devices. Needle 2 utilizes a Simple Attention Network architecture and achieves impressive speeds, even on low-power hardware like the Raspberry Pi 5. AI

IMPACT Enables advanced AI capabilities on low-power devices and microcontrollers, potentially expanding edge AI applications.

RANK_REASON Open-source model release from a known lab (Cactus Compute) with detailed technical specifications and performance metrics. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on Hugging Face Trending Models →

• Apple FM
• Apple Vision Pro
• arXiv:2607.18363
• Cactus Compute
• Cactus Quants
• ESP32-P4
• ESP32-S3
• FunctionGemma
• Meta Quest 3S
• Needle2
• Python
• Raspberry Pi 5
• Samsung A-Series
• Simple Attention Network

• model release
• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →