LT LLM Test Bench

Open-source creative model evaluation

Generate it. Compare it. Keep the evidence.

A browser-based workbench for testing how language models turn the same idea into interactive HTML, p5.js sketches, and SVG artwork.

Open Prompt Gallery Connect folder + local models How it works
Run your own local models Files stay in your chosen folder No account required

First public release

Three focused test benches

Start with 149 bundled prompt ideas, then save your own results append-only to a folder you control.

HTML + JavaScript

Prompt Gallery

Run one prompt through different models, preview the generated page safely, compare variants, refine failures, and record a walkthrough.

  • 62 starter prompts
  • Streaming generation and batch runs
  • Gallery, comparison, ratings, and metadata
Launch Prompt Gallery
Creative coding

p5 Sketch Gallery

Generate instance-mode p5.js sketches, tune their parameters and seeds, inspect lineages, compare runs, and capture video.

  • 24 ambitious starter prompts
  • Sandboxed live canvas preview
  • Remix, explain, compare, and record
Launch p5 Sketch Gallery
Vector evaluation

SVG Generation Gallery

Generate SVG submissions, compare them with reference images, inspect structure, score visual similarity, and export results.

  • 63 starter prompts
  • Pixel-diff and manual scoring
  • SVG, GIF, comparison, and recording
Launch SVG bench

Your data, your models

A local-first website

The site is hosted, but your working files are not uploaded to us. In Chrome or Edge, choose a folder on your own computer. The apps read and write inside that folder with your permission.

  1. Choose a folderThe browser remembers the handle and asks before restoring access.
  2. Connect a modelUse your local LM Studio/OpenAI-compatible server or your own OpenRouter key.
  3. Generate and comparePrompts go only to the provider you select; saved outputs remain in your folder.