Cactus Compute · Tiny Device Model

Needle 3

Since Week 33 2026, when this archive covered Needle 2. Needle 3 is an automation model for tiny devices at 29 to 121M parameters using laddered simple attention networks with CQ2 quantisation, packaged at 8 to 29MB for mobile, wearables, robots, smart home, automotive and microcontrollers. The new mechanism is intelligence laddering: one set of weights yields a usable model at any depth from 2 to 20 layers. Cactus claims a 4-layer tuned version passes DeepSeek V4 Flash on structured tasks and beats models ten times its size on mobile tool calls, at 400 to 4,000 tokens a second decode on a Raspberry Pi 5. No licence or release date stated on the page.

Open the source ↗

Similar Developer Tools releases

  • edge0

    Edge0-AI · Streaming MoE Inference · Week 37, 2026

  • Magnitude

    magnitudedev · Local Model Server · Week 36, 2026

  • SwarmLLM

    SwarmLLM · Peer-to-Peer Inference · Week 36, 2026

  • Jalapeño Inference Chip

    OpenAI / Broadcom · Inference Engine · Week 26, 2026

  • Unsloth GLM-5.2 Guide

    Unsloth · Edge Deployment · Week 25, 2026

  • Odysseus

    pewdiepie-archdaemon · Inference Engine · Week 23, 2026

More from Cactus Compute

  • Whistle

    Cactus Compute · Tiny On-Device ASR · Week 41, 2026

  • Needle

    Cactus Compute · Evaluation & Benchmarks · Week 20, 2026

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.