IndexN°07

evaLLM

One prompt, four models, the answers side by side.

Paused · not deployedHubWeb, with Expo apps for iOS and Android

SoftwarePWAMobile

01What it is

A benchmarking cockpit for indie developers and AI engineers. One prompt runs against up to four models at once and the answers render by kind: markdown for text, a live preview for code, full resolution for images. A short rubric of what a good answer must do arrives first.

Designed for acquisition, like WALL-E. It is paused until the thesis is tested with real users.

02Specification

Status
Paused · not deployed
Model access
Bring your own OpenRouter key, encrypted server-side
Planned tiers
Free · $5 a month · $20 a month

03How it works

  1. Your key, your bill

    Every call runs on the user's own OpenRouter key, so the product carries no inference cost and a $5 tier can exist.

  2. The rubric comes first

    Before any answer arrives, a model-neutral rubric of 150 words or fewer says what success looks like.

  3. Rendered by kind

    Text as markdown, code as a live preview, images at full resolution, side by side.