Skip to content

Free browser tools that respect your files: everything runs locally, nothing is uploaded.

Prompt Playground

Fill the variables, run the template across models side by side — cost and latency per column.

Free trial — run 1 of 3 today

Template

≈ 1 tokens (estimate)

Models

Trial runs one model

Temperature and token controls run with provider defaults; per-run overrides are on the roadmap.

per model, from list prices

Results

Columns appear here — one per selected model, streaming as they answer, each with its own latency, tokens, and estimated cost. Fill the variables and press Run.

Quick Answer

A free prompt playground that runs in your browser with your own API key: write a template with {{variables}} (typed — text, number, date, select, long), fill the auto-generated form, set temperature and max tokens, and run it across up to 3 models side by side. Each column streams its answer with live latency, token counts, and cost estimated from provider prices; every run is logged locally so you can compare later. 3 free trial runs a day on this site’s key (single model); unlimited multi-model runs with your own OpenRouter, Anthropic, Gemini, or Ollama key.

How It Works

1

Write or paste a template with {{variables}} — the Variables panel fills with one typed field per variable (press Fill example to see it work)

2

Pick up to 3 models from your active provider (with an OpenRouter key that can be GPT + Claude + Gemini in one run)

3

Press Run: one streaming column per model appears with its own latency, token counts, and estimated cost

4

Compare the answers, copy the best, and check the Run log for cost history — every run is recorded locally

Key Facts

  • {{variable}} templates with typed fields — {{name|number}}, {{name|date}}, {{name|select:a|b|c}}, {{name|long}}
  • Run one prompt across up to 3 models from your active provider at once (an OpenRouter key covers GPT, Claude, Gemini, Llama, DeepSeek, and more)
  • Temperature and max-tokens controls with tabular readouts
  • Per-column live latency, token counts, and cost estimates from provider list prices
  • Every run appended to a local run log (model, tokens, cost, latency)
  • One-click "Fill example" for every variable; values persist per browser
  • No chat shell — a compare bench, not a conversation

Frequently Asked Questions

Why can I only compare models from one provider at a time?

Your key lives in a single Web Worker that holds one active provider key. With an OpenRouter key that is barely a limitation — one key covers GPT, Claude, Gemini, Llama, DeepSeek, Grok, and dozens more, so the 3 columns can still be three different flagship models.

How do variables work?

Write {{name}} anywhere in the prompt. The form below the editor fills with one field per variable. Type them with a pipe: {{count|number}}, {{start|date}}, {{tone|select:formal|casual|playful}}, {{notes|long}}. Values stay in this browser for your next visit.

Are the cost numbers exact?

They are estimates: token counts come from the provider’s usage report when it supplies one, otherwise from a chars-per-token heuristic, multiplied by the provider’s published list price for the model. Treat them as a close estimate, not an invoice.

Is the free trial multi-model?

The trial runs a single model (this site’s Gemini) so one trial run stays one provider call. For true side-by-side comparison, add your own key — a free Google AI Studio or OpenRouter key keeps cost at zero.