Every test
Same rules every time. Here are the tests.
Each test gets the same prompts for every model, every effort level, judges who never see a name, and the real API bill. Open one to see the numbers, the recordings and the files.
October 2026 · full page
Gemini 4 Argon vs four rivals
Ten tests, five models, three effort levels. Play every output live and judge the visual ones yourself.
150 runs5 models9.64 winning score
October 2026 · on the home page
One burning match, 21 settings
Four models at every effort level they offer drew the same burning match. The most expensive setting did not win.
84 runs21 settings$13.40 total bill
Coming next
You pick the next one.
Vote below. The most wanted test goes into the queue.
Your call