Needle 2 compact on-device tool-calling model
A 45-million-parameter model packaged as a 14 MB binary for tool calling, device control, and structured extraction on small devices.
From ManuAGI - AutoGPT Tutorials — Top Open Source GitHub Projects Sonarr, Dokku, neko, Spec Kit, WeKnora & Public APIs #29 at 08:30
Problem: Running structured tool-calling models on devices with limited memory and connectivity.
For: Developers building assistants for phones, wearables, robots, and other memory-constrained devices.
Products from this video
ArmorPaint Armory Engine awesome-gpt-image-2 BankMCP Bot Crossing Claude Code Context Mode Cursor Dokku Enable Banking Feishu Figma GitHub Copilot GitHub Spec Kit Hyperresearch i-have-adhd It's Giving Kodi MediaPipe Modular Platform Needle 2 Notion NZBGet OKF Agent Memory OpenAI Codex OpenAI GPT Image 2 OpenAI Plugins Playwright Plex Public APIs retrobios SABnzbd Sonarr SQLite WeKnora
Examples
- Needle 2: runs a full session in about 28 MB of RAM and is intended for phones, wearables, and robots.