AI Horde: free volunteer GPU network for image and LLM generation

My notes

Summary

A curated directory of client apps, worker software, bots, plugins and SDKs built on top of AI Horde, a free crowdsourced compute network for AI image generation and LLM inference. Anyone can donate spare GPU time as a “worker” and earn kudos, or use the network for free via one of dozens of third-party clients.

Key Insight

• AI Horde is a decentralized alternative to paying for Midjourney, OpenAI or Stability API access: volunteers run “workers” (GPUs) that serve requests from anyone on the network, paid for in a reputation currency (“kudos”) rather than money.
• Scale signal: Artbot alone has generated over 39 million images through the horde, and SillyTavern (the dominant LLM roleplay frontend) has 300+ contributors and 3+ years of active development. This is a mature, actively maintained ecosystem, not an abandoned hobby project.
• No GPU required to participate as a user: several clients (Lucid Creations, HordeNG, HordeQT) run natively on Windows, Linux and macOS or as PWAs and just proxy requests to the horde, which is useful for testing image and LLM generation on machines without dedicated GPUs.
• Coverage extends well past chat UIs into production and creative tooling: plugins exist for GIMP, Krita, Blender, Photoshop, Unreal Engine, Godot, LibreOffice and Foundry VTT, plus an OpenAI-compatible proxy that lets any OpenAI-SDK-based tool point at the free horde instead of a paid API.
• llmfit (listed under Utilities) is a standalone terminal tool worth noting even outside the horde context: it detects the host hardware and scores which local LLMs will actually run well on quality, speed, fit and context, useful for anyone evaluating local-inference feasibility on a given machine.