Codex 里用 Astra 两天撞上周额度上限

Dillon Loomis @DillonLoomis原推发布于 2026.09.22

在 X 上打开原推

The reason Grok 4.7 is important isn't because of any benchmark score - that's all irrelevant to the average person I've been using Grok 4.6 in Cursor for most of my work the past few weeks on the Ultra plan ($200/month). I've built 4 websites, maintained 3 dashboards, done nearly all of my daily research and a handful of other projects with it and never once got close to hitting any usage limits In the past two weeks I wanted to try GPT 6 Astra inside Codex. I'm on the Pro plan 5x ($100/month) because the 20x ($200/month) isn't available. I literally just edited a website for a weekend and ran ~4 research projects. I hit my weekly usage limit in two days You can argue all you want about proper use case for a model, but 95% of people will not need extreme capability from these models, they just want something fast, accurate and efficient. So far, Grok 4.6 has been excellent in this regard for me. Excited to see what 4.7 can do