A complete promptYour task and reusable details combined into one final request.
LIVE MULTI-MODEL LAB / REAL INFERENCEOne prompt.
One prompt.
Real model answers.
Test the same task across real frontier, open, and small models—side by side.
Build the prompt, select two to four models, run live inference, compare outputs, latency, tokens, and cost, then export the winning configuration.
WHAT YOU GETOne fair test.
One fair test.
Multiple real answers.
Live side-by-side answersThe identical request runs across every selected model.
Measured evidence + exportCompare latency, tokens, cost, output quality, and copy the winner.
Pick the outcome closest to the work you need completed.
Replace the example values with your model, audience, workload, campaign, or document.
Select two to four models, output style, creativity level, and maximum response size.
Compare real answers and measured usage, select the best result, then copy its runnable code.
What do you want AI to help you produce?
Select one of the four real-world goals below. Every field can be edited after you choose.
Sets the role, standards, and rules shared by every tested model.
The exact job each selected model receives, so the comparison stays fair.
STEP 3 / SELECT + RUN REAL MODELS
Send the same request to two to four models.
Responses run independently through Netlify AI Gateway. The model names, outputs, time, token usage, and cost are returned from the live requests.
MAXIMUM ESTIMATED TOTAL$0.00503 models · if every model uses the full answer limit
MODEL CONNECTIONUSE YOUR OPENROUTER KEY
This Netlify project does not currently provide shared AI credits. Enter your own OpenRouter key to run real models; it stays only in this browser tab and is never saved.
READY TO TEST
Gemini 3.7 Flash vs DeepSeek V4 Pro vs Nemotron 3.5 Lightning
STEP 4 / REVIEW + EXPORT THE WINNER
CHECK BENCHMARK EVIDENCE ↗Choose the strongest result, then export it.
Run the test and mark a best answer. Until then, export uses the first selected model.
FINAL PROMPT PREVIEWThe same completed text is sent to every model in the live comparison.
MODEL ROLE + RULES
You are a technical AI editor. Separate confirmed facts from interpretation and preserve source attribution.
YOUR COMPLETED TASKCreate a concise release brief about the selected model for a technical product team. Cover what changed, strongest evidence, limitations, pricing implications, and the next validation step.
EXPORT THE WINNING CONFIGURATION
Run it in your own project.
Choose a language and copy the request. Keep provider keys in environment variables—never in browser code or source control.
const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
method: "POST",
headers: {
"Authorization": `Bearer ${process.env.OPENROUTER_API_KEY}`,
"Content-Type": "application/json"
},
body: JSON.stringify({
"model": "google/gemini-3.7-flash",
"messages": [
{
"role": "system",
"content": "You are a technical AI editor. Separate confirmed facts from interpretation and preserve source attribution."
},
{
"role": "user",
"content": "Create a concise release brief about the selected model for a technical product team. Cover what changed, strongest evidence, limitations, pricing implications, and the next validation step."
}
],
"temperature": 0.3,
"max_tokens": 900
})
});
console.log(await response.json());LIVE TESTING + PRIVACY
Running sends the completed prompt to the selected model providers. Site-gateway requests use Netlify AI Gateway; visitor-key requests pass through the protected function to OpenRouter. Keys are held in component memory only and are never written to storage or returned by the server. Do not submit secrets or personal data. Model output can be inaccurate and must be reviewed.