LLMTest

Went through the flow after seeing it. The homepage is clean, the 3-step flow and pricing section do their job. But the signup page is off, full black background, no nav, no reminder of what you're getting into and there is no "no credit card needed" line. You go from a well-designed homepage to a form that feels like a different product. That moment right before someone types their email is where a lot of people reconsider. Small fix, but the signup page needs to carry some of the homepage's momentum into the form. The product itself looks solid.

The automatic fallback part is the strongest hook for me. Choosing the right model is useful, but making sure an AI feature does not break when a provider times out or returns bad JSON feels even more important in production.

Hey PH!

I'm Tom and I created LLMTest to solve one of my own problems.

I was constantly trying to figure out which LLM models to choose to power AI features in my products.

So I was using OpenRouter and Claude to create custom webapps for each use case.

But all the tests were still "manual" and I had to decide for myself which of 300+ models I would end up using.

I thought others might be facing the same issue.

So I created LLMTest: an API and MCP server that lets anyone (coder or vibe coder) just run automatic tests on their own AI flows and pick winners. And I added automatic fallbacks so that even if an API is down or not returning a JSON like requested, my apps don't break in production.

To sum things up:
✅ One API. One invoice. 300+ models available, refreshed daily.
🤖 Ask Claude Code / Codex to optimize AI features using LLMTest
🏆 Use better, faster and/or cheaper models in your apps
↪️ Smart fallbacks: forget timeouts, overloads, and bad JSON output
💰 Pay per use. Start with $5. Top-up anytime.

I'd love to get your feedback on this ❤️