Starting in December, OpenAI will pool ChatGPT and Codex usage for Plus, Pro, and Business subscribers. Chat using Instant, the lowest effort setting, will be exempt. As it stands, Instant thinking effort will limit the scope of ChatGPT usability; whether OpenAI is planning for a separate model just for “Instant”, I do not know. As someone keeping a GPT subscription alongside Claude, I have one less reason to maintain both.

OpenAI has been vocal about bringing Work’s agentic capabilities into the main ChatGPT experience. What makes this merge particularly nasty for research is that both GPT and Claude tend to drain usage without a good way to estimate how much a task will consume. That was ChatGPT’s strong suit in my workflow: I could keep research in chat without draining the Codex allowance.

It is a standard practice shared in AI user communities to keep the long-running project work in the CLI. Someone using Codex to manage projects can still have questions that need a considered answer without needing an automation job. If one is waiting for the 5-hour window to reset, then it would be a great time to do more research and investigation. Under the announced arrangement, choosing a higher thinking setting for that research will also mean using allowance that could have gone to Codex.

A cheaper model would not be enough to make me keep the subscription either. I am paying for the reasoning and judgment needed to coordinate and orchestrate different jobs. If OpenAI’s sole value proposition is keeping the price down, its lineup has become less compelling for what I need, and, in all likelihood, GPT-6 at the Instant setting would only perform as well as the free tier of Gemini.

I had already spent weeks asking OpenAI why a newly paid Plus period began with my weekly Codex allowance at zero. Support never established a definitive rule for what had happened. OpenAI somehow still does not have a proper communication channel outside of one employee’s X account (formerly Twitter) to announce bonus resets, when said resets are so frequent that they happen at least once a month.

I have considered keeping two Claude subscriptions instead of one Claude and one OpenAI subscription. What holds me back is redundancy: if Claude has a problem, I need a different agent. OpenAI still serves that purpose. But calling Instant exempt is corporate shorthand for withdrawing higher-effort chat as a separate subscription benefit.