ollama-grid-search

Why we included this project

When you have a shortlist of Ollama-served models and want to see, side by side, which one actually behaves best for your prompt, this desktop app spares you a lot of manual switching. It turns a single prompt into a matrix of runs across several models, parameter values, and repeated iterations, then lays the responses out so you can spot differences without re-running anything by hand. The A/B testing mode helps small teams compare prompts or system prompts under otherwise identical settings, and any experiment can be saved and rerun later with tweaked parameters. It is a practical, low-ceremony tool for regular informal benchmarking of local models, whether you point it at a local Ollama install or a remote server. The bundled prompt archive with slash-autocomplete and exportable JSON logs make repeat comparisons easier to share with teammates.