GPT 5 Rate Limits Decreased!

Hello,

I was quite happy to find the GPT 5 models available on launch day and managed to make them available to the company. However, the following day we started to notice the service degrading and then I was shocked to find that all models have had their rate limits capped to 20K a minute, down from 1M.

I deployed in the Sweden Central region using Data Zone Standard which worked well on launch day. Now all deployment types have very low rate limits.

When will you properly restore the GPT 5 service? Those token limits will not cut it anymore with the prevalent use of MCP tools, RAG and web searches that we have going on.