which-llm
Picking an LLM from memory is a great way to recommend last month's model at this month's price. I built which-llm so an agent can answer from current evidence instead.
The skill joins Artificial Analysis benchmarks with OpenRouter availability, prices, context windows, modalities, and provider slugs, then packs that into a checked-in snapshot. No API key is needed, and the first query works offline.
I kept the interface atomic: pick ranks candidates, compare puts named models side by side, profile explains one model, resolve finds provider slugs, and frontier plots the quality-cost or quality-speed tradeoff. The output is evidence, not a magic answer, so the agent still has to state the objective and explain the tradeoff.
A daily GitHub Action refreshes the data and commits changed snapshots. Each snapshot carries the upstream timestamp, and stale or undated data stops recommendations instead of quietly pretending old numbers are current. That fail-closed bit matters more than another fancy ranking formula.