got tired of reading comparisons that were obvisously written by people who tested each tool for 20 minutes so i ran both at $20/month for 30 days on the same tasks
biggest surprises:
- chatgpt gives you roughly 6x more messages per day at the same price
- claude wins 67% of blind code quality tests against codex
- neither one is less sycophantic than the other (stanford tested 11 models, all of them agree with you 49% more than humans do)
- the $100 tier showdown between openais new pro 5x and claudes max 5x is where the real competition is happening now
full complete deep-dive with benchmark data, claude code vs codex and every pricing tier compared here
[link] [comments]