Back to news
Large Language Models
2d ago

Cactus releases Needle 2, a compact LLM for various devices

Aug 10, 2026
AI Summary

Cactus has launched Needle 2, a 14MB agentic language model designed for phones, wearables, and smart home devices. The model operates efficiently on low-resource hardware, achieving high decoding speeds while consuming significantly less power compared to traditional models.

  • Needle 2 is a 14MB language model designed for use in mobile devices, wearables, smart homes, and small robots.
  • It operates with 45 million parameters at 2-bit compression and requires only 28MB of RAM for a full session.
  • The model achieves decoding speeds of 500 tokens per second on Raspberry Pi 5, and between 300-1,500 tokens per second on various VR devices and budget smartphones.
  • Needle 2 is designed to perform structured extraction and can be fine-tuned for specific tasks using a Python package.
  • The model is significantly smaller and more efficient than comparable models, consuming 7x to 85x fewer resources per token.
  • It incorporates a confidence scoring system to determine when to act or escalate tasks to more powerful models.
  • Users are encouraged to test Needle 2 and provide feedback to improve its functionality.
llmagenticsmart homewearablesraspberry pi

Related Stories

Large Language Models

Grok 4.6 Achieves 61 on Artificial Analysis Intelligence Index, Enhancing SpaceXAI's Position

21h ago

Large Language Models

Exploring the Mathematical Capabilities of Large Language Models

1d ago

Large Language Models

Google's Gemini app reaches over 1 billion monthly active users

1d ago

Large Language Models

Anthropic AI model advances understanding of the Riemann hypothesis

1d ago