∑LLM FitLab

Ollama model leaderboard

Pick your GPU — see what fits, how fast it runs, and how to launch it with Ollama, OllaBridge or HomePilot.

Models
—
Ollama
—
Measured
—
Updated
—
Filters
—
# Model↓ Score↓ Fit↓ Speed↓ Stack

What the score measures

Hardware fit
live
Runtime speed
measured + estimated
Task quality (MMLU, pass@1…)
not collected
Score formula & caveats

30% fit + 30% capability + 25% HF momentum + 10% speed + 5% stack

Deployment readiness, not intelligence. Speed is measured only where a benchmark exists for that model × GPU pair; everything else is a bandwidth estimate and labelled as such.

Weekly stack probe

How compatibility is decided

Green means the component passed last week's end-to-end PLUGS probe at its latest release. Per-model stack badges are derived from registry metadata: OL an Ollama tag exists, OB OllaBridge serves it over OpenAI-compatible HTTP, HP HomePilot can use it as a chat model.

Comparison