BitPull vs exo, Petals and llama.cpp RPC

All of them split a large open model across several computers. BitPull adds what a group needs around the engine: friends' computers over the internet, an app that sets everything up, and credits when nobody in the group is using the model.

AspectBitPullexo, Petals and llama.cpp RPC
SetupDesktop app for macOS, Windows and Linux: it downloads the model and the engine and splits the model by itselfInstall and configure the tool on each computer, usually from the command line
Other people's computersShared groups by invitation, also over the internet through BitPull's relayMainly your own devices on one local network (exo, llama.cpp RPC), or a public swarm (Petals)
EarningThe group's model can be lent while idle; members share half of what it earnsNo built-in way to earn
From AI toolsMCP server and an OpenAI-compatible address for Claude Code, Cursor, Codex and othersOpenAI-compatible API; tools are connected by hand
Large modelsA catalogue of open models sized for groups: up to 9 computers per copy over the internet, 24 on one networkDepends on your devices and your setup

When BitPull is the right choice

  • You want to combine the computers of friends or classmates, not only your own.
  • You want an app that sets everything up, without the command line.
  • You want the group to earn credits while nobody in it is using the model.

When another tool may fit better

exo or llama.cpp RPC may fit better if you only use your own devices on one local network and prefer open-source tools you run and configure yourself.

Comparison written in November 2026. The other services change: check their websites for their current features and prices.