Local model lab · prompt benchmarks

Builds, measured side by side.

One adaptable prompt catalog, a live gallery of model outputs, and the numbers behind each run. Same prompt versions stay comparable; execution modes remain clearly separated.

✳ Prompt catalog
Benchmark gallery
Prompt cards are discovered from PROMPTS.md; add more without changing this layout.

Model comparison

Average repeat runs for one exact prompt version and execution mode.

Run snapshot

Current prompt catalog, latest matching prompt version.

Historical runs outside the current PROMPTS.md catalog/version