AI Potluck
Model components / Inference code

LocalAI

Mudler (LocalAI)

Self-hosted, OpenAI-compatible inference engine that runs LLMs, vision, image, audio, and video models on commodity hardware (CPU or any GPU). A modular multi-backend architecture (36+ backends incl. llama.cpp, vLLM, transformers, whisper.cpp, diffusers) behind one unified API, with built-in agents, RAG, and MCP.

MIT, no feature-gated core. ~47k stars. Broadest model-type coverage of the local inference engines.

Openness

5 high confidence
5.0
license
MIT(OSI)
source
public
core-gated
ungated

Fully MIT.

Adoption

3 medium confidence
3.0

~47k stars; no clean download telemetry for a self-hosted server - stars_fallback (capped at 3).

Capability

5 high confidence
5.0

Widest model-type and backend coverage of the inference-engine peers.

Unchanged since 2026-07-30 (last edited, not re-checked)