Test AI models v2.0 is hereSee what's new
Test AI models
Blog / News

Startup.Talk interviewed us: "Data is our moat"

Startup.Talk published an interview with our founder on November 18. Here is what we mean by data as a moat, and what it means for your prompts.

Marko MilojkovicFounder, Test AI models
2 min read

Key takeaways

  • Startup.Talk published a written interview with Marko Milojkovic, the founder of Test AI models, on November 18, 2026.
  • The data we mean is measurements from real model calls, with cost and time on every test and a pass result where a quality check was run.
  • Community benchmarks use anonymous totals, show no prompts or answers, and can be turned off in Settings > Privacy.

Startup.Talk published an interview with our founder, Marko Milojkovic, on November 18. It is a 13-minute read, done in writing.

Read the full interview on Startup.Talk (opens in a new tab).

Portrait of Marko Milojkovic in a dark suit with his arms crossed, on the Startup.Talk feature card.
Credit: Startup.Talk

The headline needs one clarification

The interview's headline is "Data is our moat." A sentence like that should make anyone who pastes prompts into a tool ask what data we mean, so here is the answer.

We mean measurements. Every test makes real calls to the models you pick and records what each call cost and how long it took. When you run a quality check, the result against your rule is added. Alerts repeat a test on a schedule, so the results also show what changes over time.

Our take: a public leaderboard cannot tell you how a model handles your job. Results from real tasks can, and a model company has a reason not to be the only judge of its own models.

We count totals, not your prompts

By default, your tests count toward our community benchmarks as anonymous totals: anonymized task types and winner picks. No prompts, answers, names, or organization details are shown, and you can turn this off in Settings > Privacy.

Your tests are stored in Frankfurt until you delete them, and a few features send part of your prompt to a model you did not pick. Our privacy policy says we do not train or fine-tune AI models on your prompts or results, and we do not sell your data. What each provider does with data sent through its API is set by that provider's terms. The full picture is in where your prompt goes.

The limits of an interview

An interview is our own account of the company, and a moat is something we are still building. One test still tells you only about the prompt you ran.

Run your own test

The useful part of all this is the result on your own prompt. Write what a correct answer must include, run it on the models you are considering, and compare what passes by cost.

Find the cheaper model for your own prompt
  • 10k free credits
  • No card
Run your own test

Marko

Questions

Does "data is our moat" mean you use our prompts to train models?

No. Our privacy policy says we do not use your prompts or results to train or fine-tune AI models, sell your data, or show your exact prompts publicly.

Marko MilojkovicFounder, Test AI models

Runs Eterna Creative and built Test AI models after repeated multi-provider LLM integration work for client products.

About the team