Home

Computing Provider Wiki

Swan Chain Computing Provider turns any machine into an AI inference endpoint on the Swan Chain decentralized network. Providers connect outbound via WebSocket — no public IP, no port forwarding, no SSL certs required.

Contents

• Quick Start — get a provider running in minutes
• Configuration — config.toml and models.json reference
• Inference Mode — run without a wallet or blockchain setup
• Guides
  • CLIProxyAPI — Serve GPT-5.x as a Computing Provider
  • Cydonia 24B v4.3 (AWQ) on 4× RTX 3080
  • Qwen3.8-27B (GGUF) on 2× RTX 3080 with llama.cpp
  • GLM-4.7-Flash (GGUF) on 2× RTX 3080 with llama.cpp
  • FreeToken — Qwen3.8-27B and MoE expert offload on RTX 3080
  • Benchmark: Qwen3.6-35B-A3B — FreeToken 1× RTX 3080 vs llama.cpp 2× RTX 3080
• WebSocket Protocol — message types and flow
• Troubleshooting — common errors and fixes

Quick links

• GitHub Repository
• Swan Chain
• Discord