Block just open-sourced mesh-llm, a peer-to-peer system that lets anyone pool spare GPU compute to run large open-source AI models without relying on any cloud provider.
If a model fits on your machi...
But it still doesn’t work properly. It’s slower than running similar on local ollama.