实测 Gemini 4 Argon Low 思考模式额度消耗

Vipin @vipinxtwt原推发布于 2026.10.21

在 X 上打开原推

I tried to go easy on Gemini 4 Argon. Same prompt, two different chats, using Low thinking. Prompt: Implement a concurrent LFU cache with TTL 1st attempt: 25% usage. 2nd attempt: 53% usage. Same prompt in two separate chats, and 53% of 5 hr usage is gone. Honestly, this is huge. I don't think switching between Low, Medium, or High thinking makes much difference. The usage consumption seems excessive regardless.