SelfHostMap

Local model runtimes

Run open-weight LLMs on your own hardware and expose an OpenAI-compatible API.

Quick answer

We track 3 self-hostable local model runtimes apps: Ollama, vLLM, LocalAI. Most GitHub stars: Ollama (184k). None of them publishes an official CPU/RAM figure we could use for a server-cost estimate.

Last updated

Local model runtimes: side-by-side

AppLicenseOfficial hardwareEst. server costGitHub starsStatus
OllamaMIT (Permissive)Not publishedNot estimated (no official RAM figure)184kActive
vLLMApache-2.0 (Permissive)Not publishedGPU server (not estimated)95.1kActive
LocalAIMIT (Permissive)Not publishedNot estimated (no official RAM figure)49.9kActive

All local model runtimes apps

Ollama

Active

The simplest way to download and run open-weight LLMs locally behind an API.

  • ★ 184k
  • ⟳ Oct 22, 2026
  • Permissive
  • RAM: n/a
  • Cost: n/a
Local model runtimes

vLLM

Active

High-throughput LLM inference server for GPUs with an OpenAI-compatible API.

  • ★ 95.1k
  • ⟳ Oct 22, 2026
  • Permissive
  • RAM: n/a
  • GPU server
Local model runtimes

LocalAI

Active

OpenAI-compatible API for LLMs, vision, voice and images on any hardware.

  • ★ 49.9k
  • ⟳ Oct 21, 2026
  • Permissive
  • RAM: n/a
  • Cost: n/a
Local model runtimes

Head-to-head comparisons