Gemini 4 Argon uses 13% fewer tokens than 3.8 Flash, test finds

Haider. @haider1Originally posted 2026.10.01

Open original on X

gemini 4 argon is noticeably more token-efficient than 3.8 flash still not as efficient as gpt-6.1 sol or Astra, but it uses 13% fewer tokens -- which is good because one complaint i've had with gemini models is how many tokens they burn

This post quotes another post: “Gemini 4 Argon costs $1.99 per Intelligence Index task”