Hi everyone!
I’m currently using Codex in two ways: standalone or remote on my laptop when it’s running, and Codex Cloud when my laptop is closed.
I’m on the $100/month Pro subscription, and I’m curious about the actual difference in usage limits between these two.
I’ve seen older discussions suggesting that cloud tasks consumed significantly more usage than local ones, but those comparisons were from the GPT-5.x era, before the new cloud environments introduced at DevDay 2026.
Has anyone compared local vs cloud usage with GPT-6.1 Sol/GPT-6 Astra?. How much more of the usage limits does cloud consume for similar tasks? 5-10%, 50%, or several times more?
I’m considering buying a dedicated Mac mini to keep Codex running locally, but if cloud’s additional usage is relatively small, I’d rather avoid the hardware purchase.
Any recent benchmarks or official clarification would be appreciated. Thanks!